micromachines

Review Recent Progress in the Voltage-Controlled Magnetic Anisotropy Effect and the Challenges Faced in Developing Voltage-Torque MRAM

Takayuki Nozaki 1,*, Tatsuya Yamamoto 1, Shinji Miwa 2,3 , Masahito Tsujikawa 4, Masafumi Shirai 4, Shinji Yuasa 1 and Yoshishige Suzuki 1,3

1 National Institute of Advanced Industrial Science and Technology (AIST), Spintronics Research Center, Tsukuba, Ibaraki 305-8568, Japan; [email protected] (T.Y.); [email protected] (S.Y.); [email protected] (Y.S.) 2 The Institute of Solid State Physics, The University of Tokyo, Kashiwa, Chiba 277-8531, Japan; [email protected] 3 Graduate School of Engineering Science, Osaka University, Toyonaka, Osaka 560-8531, Japan 4 Research Institute of Electrical Communication, Tohoku University, Sendai, Miyagi 980-8577, Japan; [email protected] (M.T.); [email protected] (M.S.) * Correspondence: [email protected]



Received: 24 April 2019; Accepted: 12 May 2019; Published: 15 May 2019


Abstract: The electron spin degree of freedom can provide the functionality of “nonvolatility” in electronic devices. For example, magnetoresistive random access memory (MRAM) is expected as an ideal nonvolatile working memory, with high speed response, high write endurance, and good compatibility with complementary metal-oxide-semiconductor (CMOS) technologies. However, a challenging technical issue is to reduce the operating power. With the present technology, an electrical current is required to control the direction and dynamics of the spin. This consumes high energy when compared with electric-field controlled devices, such as those that are used in the semiconductor industry. A novel approach to overcome this problem is to use the voltage-controlled magnetic anisotropy (VCMA) effect, which draws attention to the development of a new type of MRAM that is controlled by voltage (voltage-torque MRAM). This paper reviews recent progress in experimental demonstrations of the VCMA effect. First, we present an overview of the early experimental observations of the VCMA effect in all-solid state devices, and follow this with an introduction of the concept of the voltage-induced dynamic switching technique. Subsequently, we describe recent progress in understanding of physical origin of the VCMA effect. Finally, new materials research to realize a highly-efficient VCMA effect and the verification of reliable voltage-induced dynamic switching with a low write error rate are introduced, followed by a discussion of the technical challenges that will be encountered in the future development of voltage-torque MRAM.

Keywords: voltage-controlled magnetic anisotropy; magnetoresistive random access memory; magnetic tunnel junction

1. Introduction The evolving information society has triggered the rapid spread of advanced technologies, such as Artificial Intelligence (AI), Advanced Safety Vehicle (ASV), and IoT (Internet of Things), and this has led to further industrial innovation. In the society of the future, Big-Data collected from physical space will be stored and analyzed in cyber space, which creates new social values. Such a data-driven society can only be sustained by the high-speed processing of Big-Data; therefore, reducing the power

Micromachines 2019, 10, 327; doi:10.3390/mi10050327 www.mdpi.com/journal/micromachines Micromachines 2019, 10, 327 2 of 31 consumption of nano-electronic devices is becoming increasingly crucial. One promising approach is the introduction of nonvolatile computation. It is expected that the stand-by power of future computing systems will be reduced by utilizing the nonvolatile features of spintronic devices, such as a magnetoresistive random-access memory (MRAM) while using magnetic tunnel junctions (MTJ). An MTJ consists of two ferromagnetic layers that are separated by an ultrathin insulating layer, such as magnesium oxide (MgO) [1,2]. Electrons can tunnel through the barrier when a bias voltage is applied between the two ferromagnetic layers due to the ultrathin thickness of the insulating layer. The amplitude of the tunneling current depends on the relative angle between the magnetizations in each ferromagnetic layer through a spin-dependent tunneling process, which is called the tunnel magnetoresistance (TMR) effect. The direction of the magnetizations of one of the ferromagnetic layers is fixed (reference layer), typically by exchange coupling with an antiferromagnetic material. An external field (free layer), using an electric-current, can control the direction in the other, as discussed below. In this way information is written to the memory device. Then, the information can be stored by controlling the magnetization configuration between parallel and anti-parallel states, exhibiting two resistance states, in a nonvolatile manner. MRAM has great potential to be a fast, high write endurance, and CMOS-compatible nonvolatile memory, which is suitable for embedded as well as standalone memory applications. However, one of the significant remaining challenges is to reduce the energy that is needed to write information, that is, to switch the magnetization. In the long history of magnetism, magnetic fields that are produced by electric-current have been used for magnetization reversal. This indirect approach is inefficient and not scalable. Spintronics has brought us a new way of switching the magnetization through the s-d exchange interaction between the conduction electron spin and localized spin, called the spin-transfer torque (STT) effect [3–8]. The spin angular momentum that is carried by conduction electrons can be transferred to localized electrons and can induce magnetization reversal. Recently, an alternative technique for magnetization switching using the spin Hall effect, which is called the spin-orbit torque (SOT) switching [9–12], has also been attracting attention. A typical SOT device comprises a bilayer that consists of a non-magnetic heavy metal layer, such as Ta or W, and a ferromagnetic layer capped by an oxide. A transverse pure-spin current is generated when an in-plane electric-current is injected into the bilayer due to the spin Hall effect. The accumulation of spin at the heavy metal/ferromagnet interface exerts a torque and induces magnetization switching. In this switching scheme, high write endurance can be realized, even with high speed switching of the order of a few nanoseconds, because the read and write passes are separate. STT-based switching (STT-MRAM) has brought a drastic reduction in writing energy and expanded potential for applications; STT-MRAM [13–15]. Figure1 summarizes the reported writing energies for a MRAM (red dots) and STT-MRAM (blue dots) as a function of the MTJ cell size. For example, recent developments in STT-MRAMs have achieved writing energies of approximately 100 fJ/bit in perpendicularly magnetized MTJs [13], which is close to the writing energy for a dynamic-RAM (DRAM). However, it is still much higher than that of a static-RAM (SRAM), which is made up of several MOSFETs that an electric-field operates. Furthermore, a writing energy of 100 fJ/bit corresponds 7 to 10 kBT (kB is the Boltzmann constant and T is the temperature, assumed to be 300 K). On the other hand, the energy that is required to maintain magnetic information, i.e. the thermal stability, is about 60 kBT (green line in Figure1), which means that we have a large energy gap between data writing and retention, in the order of 105. This difference mainly comes from unwanted energy consumption due to ohmic dissipation of the electric-current flow. Overcoming this fundamental issue using a novel way of electric-field based spin manipulation is strongly desired. Not only for MRAMs, but all of the nonvolatile memories that have been proposed so far have a dilemma of choosing between stable nonvolatility and high operating energy. Therefore, the development of a novel type of memory having low operating energy as well as low stand-by energy can have great impact on the design of future memory hierarchy. Micromachines 2019, 10, 327 3 of 31 Micromachines 2019, 10, x 3 of 31

Figure 1. Reported writing energy for toggle magnetoresistive random-access memory (MRAM) Figure 1. Reported writing energy for toggle magnetoresistive random-access memory (MRAM) (red (red dots) and spin-transfer torque-based switching (STT-MRAM) (blue dots) as a function of magnetic dots) and spin-transfer torque-based switching (STT-MRAM) (blue dots) as a function of magnetic tunnel junctions (MTJ) cell size and the target area for voltage-torque MRAM. tunnel junctions (MTJ) cell size and the target area for voltage-torque MRAM. Various kinds of approaches to the electric-field manipulation of spin have been proposed and Various kinds of approaches to the electric-field manipulation of spin have been proposed and experimentally demonstrated, such as using the inverse magnetostriction effect in a multilayered experimentally demonstrated, such as using the inverse magnetostriction effect in a multilayered stack with piezoelectric materials [16–18], the gate-controlled Curie temperature in ferromagnetic stack with piezoelectric materials [16–18], the gate-controlled Curie temperature in ferromagnetic semiconductors [19–21] or even in an ultrathin ferromagnetic metal layer [22], magnetoelectric semiconductors [19–21] or even in an ultrathin ferromagnetic metal layer [22], magnetoelectric switching of exchange bias [23–26], electric polarization induced control in magnetic anisotropy at the switching of exchange bias [23–26], electric polarization induced control in magnetic anisotropy at ferromagnetic/ferroelectric interface [27,28], electric-field induced magnetic phase transition through the ferromagnetic/ferroelectric interface [27,28], electric-field induced magnetic phase transition structural phase transition [29], and the utilization of multiferroic materials [30,31]. However, each through structural phase transition [29], and the utilization of multiferroic materials [30,31]. of these approaches have the drawbacks of limited operation temperature or low write endurance However, each of these approaches have the drawbacks of limited operation temperature or low or difficulty in the introduction to magnetoresistive devices, although these requirements should be write endurance or difficulty in the introduction to magnetoresistive devices, although these simultaneously satisfied for memory applications. We have focused on the voltage-controlled magnetic requirements should be simultaneously satisfied for memory applications. We have focused on the anisotropy (VCMA) effect in an ultrathin ferromagnetic layer [32,33] to overcome this problem. voltage-controlled magnetic anisotropy (VCMA) effect in an ultrathin ferromagnetic layer [32,33] to This paper reviews recent progress in the research of the VCMA effect and the challenges that are overcome this problem. faced in developing new types of MRAM controlled by voltage, called voltage-torque MRAM (also This paper reviews recent progress in the research of the VCMA effect and the challenges that called Magnetoelectric (ME)-RAM) [34–39]. Section2 presents an overview of the early experimental are faced in developing new types of MRAM controlled by voltage, called voltage-torque MRAM observations of the VCMA effect in all-solid state devices and the concept of voltage-induced dynamic (also called Magnetoelectric (ME)-RAM) [34–39]. Section Ⅱ presents an overview of the early switching, with a discussion of the technical challenges. In Section3, the current understanding of the experimental observations of the VCMA effect in all-solid state devices and the concept of voltage- physical origin of the VCMA effect is discussed through experimental investigations while using X-ray induced dynamic switching, with a discussion of the technical challenges. In section Ⅲ, the current absorption spectroscopy (XAS) and magnetic circular dichroism (XMCD) analyses with first-principles understanding of the physical origin of the VCMA effect is discussed through experimental calculation. Section4 presents the materials research being done to enhance the VCMA e ffect, especially investigations while using X-ray absorption spectroscopy (XAS) and magnetic circular dichroism focusing on the heavy metal doping technique. Finally, in Section5, experimental demonstrations (XMCD) analyses with first-principles calculation. Section Ⅳ presents the materials research being of reliable voltage-induced dynamic switching and an understanding of the voltage-induced spin done to enhance the VCMA effect, especially focusing on the heavy metal doping technique. Finally, dynamics are discussed, together with a discussion on the theoretical investigations being made. in section Ⅴ, experimental demonstrations of reliable voltage-induced dynamic switching and an understanding2. Overview of of the the VCMA voltage-induced Effect and spin Voltage-Induced dynamics are Dynamicdiscussed, Switching together with a discussion on the theoretical investigations being made. Weisheit et al. first reported the VCMA effect in a 3d transition ferromagnetic layer in 2007 [32]. 2They. Overview observed of the a coercivity VCMA Effect change and of Voltage-Induced a few % in 2–4 nm-thick Dynamic FePt(Pd) Switching films immersed in a liquid electrolyte. Opposing trends in the change in coercivity in FePt and FePd, depending on the applied voltage,Weisheit were et observed. al. first reported An electric the doubleVCMA layereffect is in e ffa ective3d transition for applying ferromagnetic a large electric-field layer in 2007 at [32]. the Theyinterface; observed however, a coercivity the operating change speed of a is few limited % in and 2–4 wenm-thick need to FePt(Pd) take care films of the immersed influence ofin chemicala liquid electrolyte.reactions. TheOpposing voltage trends control in the of change in-plane in magnetic coercivity anisotropy in FePt and was FePd, also depending found in on ferromagnetic the applied voltage,semiconductors were observed. at low temperatureAn electric double [40]. Theoretical layer is effective attempts for toapplying understand a large the electric-field physical origin at the of interface;the VCMA however, effect in metalthe operating startedaround speed is the limited sametime. and Duanwe need et al. to proposedtake care thatof the spin-dependent influence of chemical reactions. The voltage control of in-plane magnetic anisotropy was also found in ferromagnetic semiconductors at low temperature [40]. Theoretical attempts to understand the

Micromachines 2019, 10, x 4 of 31 Micromachines 2019, 10, 327 4 of 31 physical origin of the VCMA effect in metal started around the same time. Duan et al. proposed that spin-dependent screening of the electric-field can induce modification in the surface magnetization screening of the electric-field can induce modification in the surface magnetization and magnetic and magnetic anisotropy [41]. Nakamura et al. calculated the VCMA effect in a freestanding Fe(001) anisotropy [41]. Nakamura et al. calculated the VCMA effect in a freestanding Fe(001) monolayer and monolayer and pointed out that electric-field induced changes in the band structure, especially the p pointed out that electric-field induced changes in the band structure, especially the p orbitals near the orbitals near the Fermi level, which are coupled to the d states, play an important role [42]. Tsujikawa Fermi level, which are coupled to the d states, play an important role [42]. Tsujikawa et al. studied et al. studied the VCMA effect in a Pt/Fe/Pt/vacuum system and found that relative modification in the VCMA effect in a Pt/Fe/Pt/vacuum system and found that relative modification in the electron the electron filling of the 3d orbital induced by the accumulated charges at the interface causes a filling of the 3d orbital induced by the accumulated charges at the interface causes a change in the change in the perpendicular magnetic anisotropy (PMA) [43]. Other possible mechanisms have also perpendicular magnetic anisotropy (PMA) [43]. Other possible mechanisms have also been discussed, been discussed, such as electric-field induced modification in Rashba spin-orbit anisotropy [44,45] such as electric-field induced modification in Rashba spin-orbit anisotropy [44,45] and electric-field and electric-field induced atomic displacement at the interface between ferromagnetic oxide and induced atomic displacement at the interface between ferromagnetic oxide and dielectric layers [46]. dielectric layers [46]. We attempted to apply the VCMA effect in an all solid state structure, which consisted of epitaxial We attempted to apply the VCMA effect in an all solid state structure, which consisted of Au/ultrathin Fe(Co)/MgO/polyimide/ITO junctions grown on MgO(001) substrates (see Figure2a) epitaxial Au/ultrathin Fe(Co)/MgO/polyimide/ITO junctions grown on MgO(001) substrates (see to investigate the feasibility for practical applications [33,47]. Figure2b shows an example of polar Figure 2a) to investigate the feasibility for practical applications [33,47]. Figure 2b shows an example magneto-optical Kerr effect (MOKE) hysteresis curves that were measured under the application of of polar magneto-optical Kerr effect (MOKE) hysteresis curves that were measured under the a voltage. The thickness of the Fe80Co20 layer is fixed at 0.58 nm. The bias direction is defined with application of a voltage. The thickness of the Fe80Co20 layer is fixed at 0.58 nm. The bias direction is respect to the top ITO electrode. A clear change in the saturation field in the out-of-plane direction can defined with respect to the top ITO electrode. A clear change in the saturation field in the out-of- be seen, which suggests a modification in the PMA. Under the application of a positive bias, the PMA plane direction can be seen, which suggests a modification in the PMA. Under the application of a is suppressed and the in-plane anisotropy becomes more stable. On the other hand, the application of positive bias, the PMA is suppressed and the in-plane anisotropy becomes more stable. On the other a negative voltage enhances the PMA and even the transition of the magnetic easy axis can be realized hand, the application of a negative voltage enhances the PMA and even the transition of the magnetic from the in-plane to the out-of-plane direction. easy axis can be realized from the in-plane to the out-of-plane direction.

Figure 2. 2. (a)( aSchematic) Schematic illustration illustration of sample of sample stack stackused for used the for first the demonstration first demonstration of the voltage- of the voltage-controlledcontrolled magnetic magnetic anisotropy anisotropy (VCMA) (VCMA) effect in eff anect all-solid in an all-solid state structure, state structure, and ( andb) applied (b) applied bias biasvoltage voltage dependence dependence of the ofpolar-magneto-optical the polar-magneto-optical Kerr effect Kerr (MOKE) effect (MOKE) hysteresis hysteresis curves for curves a 0.58 for nm- a

0.58thick nm-thick Fe80Co20 Felayer.80Co 20 layer.

Due toto screeningscreening byby free free electrons, electrons, the the penetration penetration of of the the electric-field electric-field into into a metal a metal is limited is limited to the to surface,the surface, unlike unlike in the in the case case ofa of semiconductor; a semiconductor; however, however, if the if the thickness thickness of theof the ferromagnetic ferromagnetic layer layer is thinis thin enough, enough, e.g. e.g. several several monoatomic monoatomic layers, layers, the modulation the modulation in the electronicin the electronic structure structure at the interface at the caninterface make acan sizable make impact a sizable on the impact magnetic on properties.the magnetic Details properties. of an experimental Details of verification an experimental for the physicalverification origin for ofthe the physical VCMA origin effect of are the discussed VCMA effect in Section are discussed2. in Section Ⅱ. One great great advantage advantage of of the the VCMA VCMA effect effect is its is itshigh high applicability applicability in a inMTJ a MTJstructure, structure, which which is the ismost the important most important practical practical devices devices in spintronics. in spintronics. Figure Figure3 exhibits3 exhibits the first the demonstration first demonstration of the of the VCMA effect that was observed in a MTJ structure, which consisted of Cr/ Au/ultrathin VCMA effect that was observed in a MTJ structure, which consisted of Cr/ Au/ultrathin Fe80Co20(0.5 Fe Co (0.5 nm)/MgO(t )/Fe grown on a MgO(001) substrate [48]. Here, we made electrical nm)/MgO(80 20 tMgO)/Fe grownMgO on a MgO(001) substrate [48]. Here, we made electrical ferromagnetic ferromagnetic resonance (FMR) measurements through the TMR effect. The PMA energy, K , was resonance (FMR) measurements through the TMR effect. The PMA energy, KPMA, was evaluatedPMA from evaluatedthe resonant from frequency the resonant of the frequency free layer ofat theeach free applied layer atvoltage. each applied In addition voltage. to FMR In addition measurements, to FMR measurements,the effect of a bias the voltage effect of on a bias normalized voltage on TMR normalized curves has TMR also curves often has been also used often for been the quantitative used for the quantitativeevaluation of evaluation the VCMA of effect, the VCMA as discussed effect, as late discussedr [49]. Generally, later [49]. Generally,the PMA theenergy PMA linearly energy changes linearly changes as a function of the applied electric field, E, which is defined as the applied bias voltage, as a function of the applied electric field, E, which is defined as the applied bias voltage, Vbias, divided V , divided by the MgO thickness, t . The slope of the linear relationship represents the VCMA bybias the MgO thickness, tMgO. The slopeMgO of the linear relationship represents the VCMA coefficient in

Micromachines 2019, 10, 327 5 of 31 Micromachines 2019, 10, x 5 of 31 coefficient in units of J/Vm, e.g. 37 fJ/Vm for the case in Figure3. The VCMA coe fficient is one of units of J/Vm, e.g. −37 fJ/Vm for− the case in Figure 3. The VCMA coefficient is one of the most theimportant most important parameters parameters for demonstrating for demonstrating scalability scalability and also and alsoin the inthe reliable reliable switching switching of of the magnetization and thus the development ofof voltage-torquevoltage-torque MRAM.MRAM.

Figure 3. Example of applied electric-field dependence of KPMAtfree observed in an MgO-based MTJ Figure 3. Example of applied electric-field dependence of KPMAtfree observed in an MgO-based MTJ structure. Reprinted figure with permission from [48], Copyright 2010 by the AIP Publishing LLC. structure. Reprinted figure with permission from [48], Copyright 2010 by the AIP Publishing LLC. The realization of the VCMA effect in all-solid state devices, including a MTJ structure, made it possibleThe realization for us to demonstrateof the VCMA the effect high in speedall-solid response state devices, of this including effect, such a MTJ as in structure, voltage-induced made it ferromagneticpossible for us resonance to demonstrate excitation the [high50–54 speed], dynamic respon magnetizationse of this effect, switching such as driven in voltage-induced solely by the applicationferromagnetic of aresonance voltage [55 excitation], and spin [50–54], wave excitationdynamic magnetization [56–58]. switching driven solely by the applicationIn addition of a voltage to ultrathin [55], and epitaxial spin wave films excitation with large [56–58]. PMA [35,59–67], VCMA effects have been observedIn addition in various to ultrathin materials epitaxial systems, films for example,with large in PMA sputter-deposited [35, 59–67], VCMA CoFeB effects [68–81 ],have which been is anobserved important in various practical materials material syst thatems, is usedfor example, in the mass in sputter-depo productionsited of MTJs, CoFeB and [68–81], in self-assembled which is an nano-islandsimportant practical [82], nanocomposite material that is structures used in the [83 mass], and production ultrathin layersof MTJs, with and quantum in self-assembled well states nano- [84]. Theislands VCMA [82], enanocompositeffect can also be structures applied for[83], the and control ultrathin of domain layers with wall quantum motion [85 well–87 ]states and magnetic[84]. The skyrmionsVCMA effect [88 –can90]. also In addition, be applied voltage for the control control of theof magneticdomain wall properties motion has [85–87] been expandedand magnetic not onlyskyrmions for the [88-90]. PMA, In but addition, also for voltage the Curie control temperature of the magnetic [22], Dzyaloshinskii-Moriya properties has been expanded interactions not only [91], interlayerfor the PMA, exchange but also coupling for the [92 Curie], and temperature proximity-induced [22], Dzyaloshinskii-Moriya magnetism in non-magnetic interactions metal [91], thin filmsinterlayer [93–95 exchange]. coupling [92], and proximity-induced magnetism in non-magnetic metal thin filmsThe [93–95]. VCMA effect can induce a transition of the magnetic easy axis between the in-plane and out-of-planeThe VCMA directions effect bycan the induce application a transition of a static of th voltage;e magnetic however, easy bi-stableaxis between switching the in-plane is not easily and attained,out-of-plane because directions the VCMA by the e applicationffect does not of breaka static the voltage; time reversal however, symmetry. bi-stable Oneswitching possible is not way easily is to useattained, the VCMA because eff ectthe toVCMA assist othereffect externaldoes not fields. break Forthe example,time reversal the coercivity symmetry. of One the perpendicularlypossible way is magnetizedto use the filmVCMA can effect be reduced to assist by theother application external offields. dc voltage For example, [47,96,97] the or ofcoercivity voltage-induced of the FMRperpendicularly [98], just as magnetized in the microwave-assisted film can be reduced magnetization by the application reversal (MAMR) of dc voltage technique. [47,96,97] Moreover, or of thevoltage-induced combination ofFMR STT [98], [99, 100just] oras SOTin the [101 microwave-assisted] and the VCMA e ffmagnetizationect has also been reversal experimentally (MAMR) demonstrated.technique. Moreover, These approachesthe combination are eff ofective STT in[99, reducing100] or SOT the energy [101] and that the is required VCMA effect for writing has also by electric-currentbeen experimentally based demonstrated. manipulation; however,These approaches the realization are effective of magnetization in reducing switching the energy solely that by is a voltagerequired e ffectfor iswriting much more by preferable.electric-current based manipulation; however, the realization of magnetizationWe proposed switching pulse voltage-induced solely by a voltage dynamic effect switchingis much more to overcome preferable. this problem (see Figure4). This techniqueWe proposed was pulse first demonstrated voltage-induced in in-plane dynamic magnetized switching to MTJs overcome [55,102 ]this and problem it was then (see applied Figure in4). perpendicularly-magnetizedThis technique was first demons MTJstrated [103–109 in ].in-plane For example, magnetized we assume MTJs the [55,102] initial and state it (Figure was then4a) toapplied be the in perpendicularly perpendicularly-magnetized magnetized “up” MTJs state [103–109]. under the For application example, ofwe an assume in-plane the bias initial magnetic state (Figure 4a) to be the perpendicularly magnetized “up” state under the application of an in-plane bias field, Hbias. When a short pulse voltage is applied to eliminate the PMA completely, the magnetization magnetic field, Hbias. When a short pulse voltage is applied to eliminate the PMA completely, the starts to precess around the Hbias (Figure4b). If the voltage pulse is turned o ff at the timing of half turn magnetization starts to precess around the Hbias (Figure 4b). If the voltage pulse is turned off at the precession, then the magnetization can be stabilized in the opposite “down” direction (Figure4c). Hbias timing of half turn precession, then the magnetization can be stabilized in the opposite “down”

Micromachines 2019, 10, 327 6 of 31 Micromachines 2019, 10, x 6 of 31 directionis required (Figure to determine 4c). Hbias theis required axis of magnetization to determine the precession. axis of magnetization The effective precession. field, such The as crystalline effective field,anisotropy such as field crystalline and the anisotropy exchange bias field field, and isthe also exchange applicable. bias field, is also applicable.

Figure 4. Conceptual diagram of voltage-induced dynamic switching for a perpendicularly-magnetized Figure 4. Conceptual diagram of voltage-induced dynamic switching for a perpendicularly- film. The in-plane bias magnetic field, Hbias, which determines the axis of the precessional dynamics, magnetized film. The in-plane bias magnetic field, Hbias, which determines the axis of the precessional is applied in the +x direction. (a) initial state (point S), (b) precessional switching process induced by dynamics, is applied in the +x direction. (a) initial state (point S), (b) precessional switching process an application of pulse voltage (from point S to point M), and (c) relaxation process (from point M to induced by an application of pulse voltage (from point S to point M), and (c) relaxation process (from point E). point M to point E). Figure5a shows an example of an experimental demonstration of voltage-induced dynamic Figure 5a shows an example of an experimental demonstration of voltage-induced dynamic switching being observed in perpendicularly magnetized MTJs [105]. The top FeB layer with a W cap switching being observed in perpendicularly magnetized MTJs [105]. The top FeB layer with a W cap is the voltage-driven free layer. Under an optimized applied magnetic field, we achieved the stable is the voltage-driven free layer. Under an optimized applied magnetic field, we achieved the stable toggle switching by the successive application of voltage pulses with a width of 1 ns and amplitude of toggle switching by the successive application of voltage pulses with a width of 1 ns and amplitude 1.2 V. The precessional dynamics of the magnetization are reflected in the oscillation of the switching of− −1.2 V. The precessional dynamics of the magnetization are reflected in the oscillation of the probability (PSW) as a function of pulse width, as shown in Figure5b. A high PSW is obtained at the switching probability (PSW) as a function of pulse width, as shown in Figure 5b. A high PSW is obtained timing of half turn precession; however, when the pulse width is twice this, one turn precession results at the timing of half turn precession; however, when the pulse width is twice this, one turn precession in low PSW. From a practical point of view, the first half turn precession is effective in obtaining a results in low PSW. From a practical point of view, the first half turn precession is effective in obtaining low WER with fast switching speed. Under the condition that the PMA is completely eliminated, the a low WER with fast switching speed. Under the condition that the PMA is completely eliminated, amplitude of Hbias determines the precession frequency, and then the switching time, tSW for the half the amplitude of Hbias determines the precession frequency, and then the switching time, tSW for the turn precession is expressed as half turn precession is expressed as   π 1 α2 tSW − (1) ∼𝜋(1γµ0Hbias − 𝛼 ) 𝑡~ (1) where α, γ, and µ0 are the magnetic damping constant,𝛾𝜇𝐻 the gyromagnetic ratio, and the permeability of vacuum, respectively. where α, γ, and μ0 are the magnetic damping constant, the gyromagnetic ratio, and the permeability of vacuum,The possible respectively. advantages of voltage-induced dynamic switching are as follows. (i) Fast switching (~1 nanosecond)The possible advantages can be induced of voltage-induced with an ultralow dynamic switching switching power are as of follows. the order (ⅰ) Fast of a switching few fJ/bit. ((ii) ~1 Thenanosecond) switching can transistor be induced can be wi downsized,th an ultralow because switching we do notpower need of to the apply order a large of a electric-current.few fJ/bit. (ⅱ) The(iii) switching Unipolar switchingtransistor can can be separate downsized, the polarity because of we voltages do not need for writing to apply and a large reading. electric-current. In addition, (theⅲ) VCMA-inducedUnipolar switching enhancement can separate in the PMA polarity has been of voltages used to proposefor writing a unique and reading. approach In addition, to reduce the the VCMA-inducedread disturbance enhancement [110]. in PMA has been used to propose a unique approach to reduce the read disturbance [110].

Micromachines 2019, 10, 327 7 of 31 Micromachines 2019, 10, x 7 of 31

On the other hand, the following technical challenges remain. Firstly, the realization of a On the other hand, the following technical challenges remain. Firstly, the realization of a large large VCMA effect is the most important issue to show the scalability of the voltage-torque MRAM, VCMA effect is the most important issue to show the scalability of the voltage-torque MRAM, as as discussed in Section4. Furthermore, as seen in Figure5b, the switching probability is sensitive discussed in Section Ⅳ. Furthermore, as seen in Figure 5b, the switching probability is sensitive to to the writing pulse width, due to the precession-mediated switching process. Therefore, we need the writing pulse width, due to the precession-mediated switching process. Therefore, we need verification as to whether a sufficiently-low WER can be achieved by the voltage-induced dynamic verification as to whether a sufficiently-low WER can be achieved by the voltage-induced dynamic switching technique. In addition, this is a toggle switching technique, so pre-read and read-verify switching technique. In addition, this is a toggle switching technique, so pre-read and read-verify processes are always required for writing. These reading processes dominate the total write time, and processes are always required for writing. These reading processes dominate the total write time, and it can be critical when the resistance of the MTJ cell increases. In addition, the removal of the external it can be critical when the resistance of the MTJ cell increases. In addition, the removal of the external magnetic field is also an important issue for practical applications. magnetic field is also an important issue for practical applications.

Figure 5. Experimental demonstration of voltage-induced dynamic switching. (a) Schematic of the Figure 5. Experimental demonstration of voltage-induced dynamic switching. (a) Schematic of the sample structure of a voltage-controlled perpendicularly-magnetized MTJ and observed bi-stable sample structure of a voltage-controlled perpendicularly-magnetized MTJ and observed bi-stable switching between parallel and antiparallel magnetization configurations induced by successive pulse switching between parallel and antiparallel magnetization configurations induced by successive voltage applications. (b) Pulse width dependence of switching probability, PSW. Due to the precessional pulse voltage applications. (b) Pulse width dependence of switching probability, PSW. Due to the dynamics, PSW exhibits oscillatory behavior depending on the pulse width. precessional dynamics, PSW exhibits oscillatory behavior depending on the pulse width. 3. Physical Origin of the VCMA Effect 3. Physical Origin of the VCMA Effect In this section, recent experimental trials conducted to understand the physical origin of the VCMAIn this effect section, are introduced recent experimental [111]. The following trials cond twoucted mechanisms to understand account the for physical the purely origin electronic of the VCMA effect effect. are The introduced first mechanism [111]. The comes following from the two charge-doping-induced mechanisms account for anisotropy the purely in theelectronic orbital VCMAangular effect. momentum, The first as mechanism shown in comes Figure 6froma. As the each charge-doping-induced electron orbital in the anisotropy vicinity ofin the orbital Fermi angularlevel has momentum, a different densityas shown of in states, Figure selective 6a. As each charge electron doping orbital may inducein the vicinity anisotropy of the in Fermi the orbital level hasangular a different momentum. density Thisof states, effect selective changes charge the PMA doping energy may through induce anisotropy spin-orbit interactionsin the orbital from angular the momentum.spin-conserved This virtual effect excitation changes processes the PMA [112 energy,113], asthrough expressed spin-orbit by the firstinteractions term in Equation from the (2) spin- [114]. conserved virtual excitation processes [112, 113], as expressed by the first term in Equation (2) [114]. 1 λD E D E 7 λD E D E + 1 𝜆 ∆Lξ, ∆Lξ, 7 𝜆 ∆Tζ0, ∆Tζ0, (2) −− 4 }〈Δ𝐿 ↓↓〉 −− 〈Δ𝐿 ↑↑ 〉+2 } 〈Δ𝑇 ↓↑ 〉−− 〈Δ𝑇↑↓ 〉 (2) 4 ℏ ,↓↓ ,↑↑ 2 ℏ ,↓↑ ,↑↓ Here, λ is the spin-orbit interaction coefficient. L and T0 are the orbital angular momentum and Here, λ is the spin-orbit interaction coefficient. L and T’ are the orbital angular momentum and part of the magnetic dipole operator, respectively. Here, ∆Lξ Lz Lx and ∆Tζ0 Tz0 Tx0 part of the magnetic dipole operator, respectively. Here, 〈hΔ𝐿〉i≡ ≡〈 h𝐿〉i− −〈 h𝐿〉i and h〈Δ𝑇 〉i≡ h〈𝑇 〉i− h〈𝑇〉i are used. Lz and Lx are evaluated for the z- and x- components of the spin angular momentum, are used. 〈h𝐿〉i and h〈𝐿i〉 are evaluated for the z- and x- components of the spin angular momentum, respectively. The same is the case for Tz0 and Tx0 . and denote the contributions from the majority h 〈i𝑇〉 h i〈𝑇↑〉 ↑ ↓ ↓ respectively.and minority The spin-bands, same is the respectively. case for We and call the . first and mechanism denote the the orbital contributions magnetic from moment the majoritymechanism. and Theminority second spin-bands, mechanism respectively. is the VCMA We effect call from the thefirst induction mechanism of an the electric orbital quadrupole magnetic moment(Figure6 b).mechanism. An electric-field The second applied mechanism to the metal is the/dielectric VCMA effect interface from is the inhomogeneous, induction of an owing electric to quadrupolethe strong electrostatic (Figure 6b). screeningAn electric-field effect inapplied the metal, to the such metal/dielectric as electric-field, interface including is inhomogeneous, higher-order owing to the strong electrostatic screening effect in the metal, such as electric-field, including higher- order quadratic components, can couple with the electric quadrupole correlated with the magnetic dipole operator in an electron shell of the metal layer. The induced energy split of each orbital changes

Micromachines 2019, 10, 327 8 of 31

Micromachines 2019, 10, x 8 of 31 quadratic components, can couple with the electric quadrupole correlated with the magnetic dipole theoperator magnetic in an anisotropy electron shell through of the spin-orbit metal layer. interactions The induced from energy spin-flip split virtual of each excitation orbital changes processes the [115,116],magnetic anisotropyas shown in through Figure spin-orbit 6c. The latter interactions mechanism from spin-flipcorresponds virtual to the excitation second processes term in Equation [115,116], (2).as shown We call in this Figure the 6electricc. The latterquadrupole mechanism mechanism. corresponds As the to expectation the second values term in for Equation the orbital (2). angular We call momentumthis the electric and quadrupole the magnetic mechanism. dipole operator As the can expectation be measured values as forthe theorbital orbital magnetic angular moment momentum and and the magnetic dipole operator can be measured as the orbital magnetic moment and the magnetic the magnetic dipole Tz term (mT), respectively, the aforementioned two mechanisms can be validated bydipole X-rayTz termabsorption (mT), respectively, spectroscopy the (XAS) aforementioned and X-ray two magnetic mechanisms circular can bedichroism validated by(XMCD) X-ray spectroscopy.absorption spectroscopy (XAS) and X-ray magnetic circular dichroism (XMCD) spectroscopy.

Figure 6. Microscopic origin of the VCMA effect. (a) Orbital magnetic moment mechanism. (b) Electric Figure 6. Microscopic origin of the VCMA effect. (a) Orbital magnetic moment mechanism. (b) Electric quadrupole mechanism. (c) Schematic of the nonlinear electric field at the interface between the quadrupole mechanism. (c) Schematic of the nonlinear electric field at the interface between the dielectrics and the ferromagnet, which induces a charge redistribution-induced VCMA effect. dielectrics and the ferromagnet, which induces a charge redistribution-induced VCMA effect. The XAS/XMCD experiments provide element-specific information on the electronic structure via the opticalThe XAS/XMCD transition from experiments the core levelprovide to unoccupied element-specif statesic ininformation the valence on band. the electronic Based on thestructure use of viacircularly the optical polarized transition X-rays, from X-ray the absorptioncore level to techniques unoccupied provide states interesting in the valence features band. for Based the study on the of usemagnetic of circularly materials. polarized Figure 7X-rays, shows X-ray a schematic absorption diagram techniques of the electronic provide interesting states that arefeatures involved for the in studyan optical of magnetic transition materials. from the Figure 2p core 7 to showsd valence a schematic states, which diagram is related of the to electronic XMCD at states the L edgesthat are of involvedtransition in metals. an optical The transition dichroic signal from the directly 2p core reflects to d valence the diff states,erence which in the is density related of to the XMCD states at near the Lthe edges Fermi of leveltransition between metals. the upThe and dichroic down signal spin sub-bands.directly reflects From the the difference XMCD results in the withdensity sum-rule of the states near the Fermi level between the up and down spin sub-bands. From the XMCD results with analysis [117,118], the magnetic moments (spin magnetic moment: mS, mL, and mT) can be determined sum-rulefrom the measuredanalysis [117,118], XAS/XMCD the spectra.magnetic Here, moments the measured (spin magnetic orbital magneticmoment: momentsmS, mL, and and m magneticT) can be determined from the measured XAS/XMCD spectra. Here, the measured orbital magnetic moments dipole Tz term have the following relationships; and magnetic dipole Tz term have the following relationships;

µB( ∆L + ∆L ) ∆mL =𝜇(〈Δ𝐿h ↓↓↓↓i〉 +h 〈Δ𝐿↓↓i ↓↓,〉) and Δ𝑚 =−− } ,and (3) LD 2 E D 2 E D E D E 7∆mT = µB ∆L ∆L ℏ 7µB ∆T + ∆T /} (3) − − − − ↓↑ ↑↓ ↑↑ ↓↓ −7Δ𝑚 =−𝜇〈Δ𝐿↑↑〉 − 〈Δ𝐿↓↓〉−7𝜇(〈Δ𝑇↓↑〉 + 〈Δ𝑇↑↓〉)/ℏ

Micromachines 2019, 10, 327 9 of 31

Micromachines 2019, 10, x 9 of 31 It should be noted that the PMA energy from the spin-conserved virtual excitation process (first termIt in should Equation be noted (2)) is that related the to PMA the orbitalenergy magnetic from the momentspin-conserved and the virtual PMA energy excitation from process the spin-flip (first termvirtual in excitationEquation (2)) process is related (second to termthe orbital in Equation magnetic (2)) moment is related and to the the magnetic PMA energy dipole fromTz term.the spin- flip virtual excitation process (second term in Equation (2)) is related to the magnetic dipole Tz term.

FigureFigure 7. 7. DiagramDiagram of of the the electronic electronic states states related related to to X- X-rayray absorption absorption spectroscopy spectroscopy and and X-ray X-ray magnetic magnetic circularcircular dichroism dichroism (XAS/XM (XAS/XMCD)CD) measurements at the L-edges of transition metals.

AA FeFe/Co/Co (1(1 ML) ML)/MgO/MgO multilayer multilayer was was employed employed to see to thesee changes the changes in the orbitalin the magneticorbital magnetic moment momentin XAS/XMCD in XAS/XMCD experiments experiments [113]. The [113]. sample The stack sample is depicted stack is depicted in Figure 8ina. Figure A multilayered 8a. A multilayered structure, consisting of bcc-V(001) (30 nm)/bcc-Fe(001) (0.4 nm)/Co (0.14 nm)/MgO(001) (2 nm)/SiO (5 nm)/Cr structure, consisting of bcc-V(001) (30 nm)/bcc-Fe(001) (0.4 nm)/Co (0.14 nm)/MgO(001) (22 nm)/SiO2 (5(2 nm)/Cr nm)/Au (2 (5 nm)/Au nm), was (5 deposited nm), was ondeposited a MgO(001) on a substrate. MgO(001) Figure substrate.8b shows Figure the 8b typical shows XAS the /typicalXMCD results around the L and L edges of Co with a magnetic field of 1.9 T (θ = 20 ) to saturate the XAS/XMCD results around3 the2 L3 and L2 edges of Co with a magnetic field of 1.9 T (θ◦ = 20°) to saturate themagnetization magnetization of theof the Fe /Fe/CoCo layer. layer. The The changes changes in in the the orbital orbital magneticmagnetic momentmoment and effective effective spin spin magnetic moment (m 7m ) of Co were determined while using sum-rule analysis, and they are magnetic moment (mSS −− 7mTT) of Co were determined while using sum-rule analysis, and they are summarized in Figure8c,d. We can see that m of Co with an electric-field of 0.2 V/nm is larger than summarized in Figures 8c and 8d. We can seeL that mL of Co with an electric-− field of −0.2 V/nm is that corresponding to +0.2 V/nm. Moreover, the induced change in m with θ = 20 is larger than that larger than that corresponding to +0.2 V/nm. Moreover, the inducedL change in ◦mL with θ = 20° is largerwith θ than= 70 ◦that. The with experiment θ = 70°. demonstrates The experiment that andemonstrates orbital magnetic that momentan orbital anisotropy magnetic changemoment of (0.013 0.008)µ between the magnetization angles of θ = 20 and 70 was generated in the presence of anisotropy± changeB of (0.013 ± 0.008)μB between the magnetization◦ ◦ angles of θ = 20° and 70° was applied electric fields of 0.2 V/nm. Figure8d shows the electric-field-induced change in m 7m of generated in the presence± of applied electric fields of ±0.2 V/nm. Figure 8d shows the electric-field-S − T Co. As with m , m 7m is enhanced under the application of a negative electric-field. Moreover, the induced changeL inS m−S − T7mT of Co. As with mL, mS − 7mT is enhanced under the application of a negativeelectric-field-induced electric-field. change Moreover, in the magneticthe electric-field-induced moment is anisotropic. change In contrastin the magnetic to mT, it is moment known that is m is not sensitive to the magnetization direction. Hence, the anisotropic part of the induced change in anisotropic.S In contrast to mT, it is known that mS is not sensitive to the magnetization direction. Hence,the magnetic the anisotropic moment shouldpart of bethe attributed induced tochangemT. in the magnetic moment should be attributed to mT. As discussed in the previous section, Equation (2) can be used to analyze the VCMA effect. If we employ the spin-orbit interaction coefficient of Co, λCo = 5 meV, then the induced change in the PMA energy is estimated to be 0.039 ± 0.023 mJ/m2 when the applied electric-field is switched from +0.2 V/nm to −0.2 V/nm. Here, the experimentally obtained ΔmL = (0.017±0.010)μB was used. From the VCMA coefficient in the Fe/Co/MgO system (−82 fJ/Vm), the PMA energy change at ±0.2 V/nm is 0.03 mJ/m2, which is in good agreement with the PMA energy change that was obtained using the first term of Equation (2). From the discussion above, the change in the orbital magnetic moment anisotropy in Co seems to explain the VCMA effect. However, the impact of the change in the magnetic dipole Tz term (mT) that is shown in Figure 8d on the VCMA effect remains to be seen. In

Micromachines 2019, 10, x 10 of 31

Ref. 113, a first principles study was employed to clarify this point. As a result, the VCMA effect from the spin-flip terms (ΔE↓↑ + ΔE↑↓) is found to be negligible and that from the spin-conserved terms

(ΔE ↑↑ + ΔE ↓↓) appeared to be dominant. Therefore, the change in orbital magnetic moment is responsible for the VCMA effect. Due to the large exchange splitting for Co, the observed changes in Micromachines 2019, 10, 327 10 of 31 mT do not contribute to the VCMA effect, as described by the second term in Equation (2).

Figure 8. Voltage-induced changes to the magnetic moment of Co in the Fe/Co/MgO system. Figure 8. Voltage-induced changes to the magnetic moment of Co in the Fe/Co/MgO system. (a) (a) Schematic of the sample structure. (b) Typical XAS/XMCD results around the Co-absorption Schematic of the sample structure. (b) Typical XAS/XMCD results around the Co-absorption edges. edges. (c) Voltage-induced change to the orbital magnetic moment in Co. (d) Voltage-induced changes (cto) Voltage-induced the effective spin change magnetic to the moment orbital ( mmagnetic7m ) moment in Co. Reprinted in Co. (d) figure Voltage-induced with permission changes from to [the113 ], S − T effectiveCopyright spin 2017 magnetic by the Americanmoment ( PhysicalmS − 7m Society.T) in Co. Reprinted figure with permission from [113], Copyright 2017 by the American Physical Society. As discussed in the previous section, Equation (2) can be used to analyze the VCMA effect. If we It has been reported that the spin-orbit interaction energy from a spin-flip virtual excitation employ the spin-orbit interaction coefficient of Co, λCo = 5 meV, then the induced change in the processPMA energymakes isa estimatedsignificant to contribution be 0.039 0.023 to the mJ VCMA/m2 when effect the when applied 3d/5 electric-fieldd-layered transition is switched metals from ± are+0.2 employed V/nm to [116].0.2 V Figure/nm. Here, 9a shows the experimentally an experiment obtainedal design∆ mand= a(0.017 high-angle0.010) annularµ was used.dark-field From − L ± B scanningthe VCMA transmission coefficient electron in the Fe /microscopyCo/MgO system (HAAD ( 82F-STEM) fJ/Vm), theimage PMA of energythe device. change Figure at 0.29b shows V/nm is − ± the0.03 typical mJ/m results2, which of is the in polarization good agreement-averaged with the XAS PMA and energy its XMCD change around that the was L obtained3 and L2 energy using theedges first ofterm Pt. A of perpendicular Equation (2). From magnetic the discussion field of ±60 above, mT thewas change applied in to the saturate orbital magneticthe magnetization moment anisotropy of FePt. Figuresin Co seems9c and to 9d explain show theelectric-field-induced VCMA effect. However, changes the in impact the magnetic of the change moments in the of magneticPt. The results dipole confirm a clear bias voltage inductions of mS − 7mT, while there is no significant change to mL under Tz term (mT) that is shown in Figure8d on the VCMA e ffect remains to be seen. In Ref. 113, a first voltageprinciples applications. study was employed to clarify this point. As a result, the VCMA effect from the spin-flip termsIn general, (∆E + in∆E low-symmetry) is found to systems, be negligible such and as interfaces, that from the the atomic spin-conserved electron orbital terms (may∆E possess+ ∆E ) ↓↑ ↑↓ ↑↑ ↓↓ anappeared electric quadrupole to be dominant. moment. Therefore, If the atom the change is also inspin-polarized, orbital magnetic the momentelectric quadrupole is responsible moment for the induces the anisotropic part of the spin-density distribution, i.e., the magnetic dipole Tz term (mT) VCMA effect. Due to the large exchange splitting for Co, the observed changes in mT do not contribute [114–116,118].to the VCMA In eff contrastect, as described to mT, m byS is the not second sensitive term to in the Equation magnetization (2). direction. In Ref. 116, the voltage-inducedIt has been change reported in thatmS − the7mT spin-orbit shows large interaction magnetization energy direction from a spin-flip dependence. virtual Thus, excitation the observationsprocess makes indicate a significant the significant contribution induction to the of VCMAmT in Pt e ffbyect an when external 3d/5 dvoltage.-layered A transition first-principles metals studyare employed was also conducted [116]. Figure for9 athe shows FePt/MgO an experimental system, similar design to andthe aFe/Co/MgO high-angle study. annular As dark-field a result, firstly,scanning the monoatomic transmission Pt electron layer at microscopy the interface (HAADF-STEM) with MgO makes image the ofdominant the device. contribution Figure9b to shows the VCMA effect. Moreover, while the VCMA effect from the spin-conserved terms (ΔE↑↑ + ΔE↓↓) the typical results of the polarization-averaged XAS and its XMCD around the L3 and L2 energy decreasesedges of the Pt. PMA A perpendicular energy, the magneticVCMA effect field that of is60 induced mT was by applied the applied to saturate voltage the from magnetization the spin-flip of ± termsFePt. of Figure interfacial9c,d show Pt increases electric-field-induced the PMA energy changes (ΔE ↓↑ in the+ Δ magneticE ↑↓). The moments total PMA of Pt.energy The in results the FePt/MgOconfirm a system clear bias increases voltage under inductions the condition of m 7ofm el,ectron while depletion there is no at significant the Pt/MgO change interface, to m asunder the S − T L PMAvoltage energy applications. increase by the spin-flip terms is greater than the PMA energy decrease by the spin- conservedIn general, terms. in low-symmetry systems, such as interfaces, the atomic electron orbital may possess an electric quadrupole moment. If the atom is also spin-polarized, the electric quadrupole moment induces the anisotropic part of the spin-density distribution, i.e., the magnetic dipole Tz term (mT)[114–116,118]. In contrast to mT, mS is not sensitive to the magnetization direction. In Ref. 116, the voltage-induced change in m 7m shows large magnetization direction dependence. Thus, the observations indicate S − T the significant induction of mT in Pt by an external voltage. A first-principles study was also conducted for the FePt/MgO system, similar to the Fe/Co/MgO study. As a result, firstly, the monoatomic Pt layer at the interface with MgO makes the dominant contribution to the VCMA effect. Moreover, while the VCMA effect from the spin-conserved terms (∆E + ∆E ) decreases the PMA energy, the VCMA ↑↑ ↓↓ Micromachines 2019, 10, 327 11 of 31 effect that is induced by the applied voltage from the spin-flip terms of interfacial Pt increases the PMA energy (∆E + ∆E ). The total PMA energy in the FePt/MgO system increases under the condition of ↓↑ ↑↓ electron depletion at the Pt/MgO interface, as the PMA energy increase by the spin-flip terms is greater Micromachinesthan the PMA 2019 energy, 10, x decrease by the spin-conserved terms. 11 of 31

Figure 9. 9. Voltage-inducedVoltage-induced changes changes to tothe the magnetic magnetic moment moment of Pt of in Pt the in theFe/Pt/MgO Fe/Pt/MgO system. system. (a) Schematic(a) Schematic of the of the sample sample structure structure and and its its high high-angle-angle annular annular dark-field dark-field scanning scanning transmission electron microscopymicroscopy (HAADF-STEM) (HAADF-STEM) image. image. (b) Typical(b) Typical XAS/ XMCDXAS/XMCD results aroundresults thearound Pt-absorption the Pt- absorptionedges. (c) Voltage-induced edges. (c) Voltage-induced change to the change orbital magneticto the orbital moment magnetic in Pt. (momentd) Voltage-induced in Pt. (d) Voltage- changes to the effective spin magnetic moment (m 7m ) in Pt. Reproduced from [116]. CC BY 4.0. induced changes to the effective spin magneticS − momentT (mS − 7mT) in Pt. Reproduced from [116]. CC BY 4.0. To conclude, for the 3d-transition ferromagnetic metals, it is important to consider the orbital magneticTo conclude, moment for anisotropy. the 3d-transition The validity ferromagnetic of the Bruno metals, model it is [112 important] (first term to consider of Equation the (2)orbital and d d magneticFigure6a) moment has been experimentallyanisotropy. The demonstrated validity of the in Bruno Ref. 113. model For the[112] 3 (first/5 -multilayered term of Equation ferromagnetic (2) and d Figuremetals, the6a) orbitalhas been magnetic experimentally moment anisotropy demonstrated in 3 -metals in Ref. cannot 113. completelyFor the 3 explaind/5d-multilayered the VCMA d d ferromagneticeffect. In addition metals, to the the magnetic orbital magnetic moments moment in 3 metals, anisotropy those in in 5 3metalsd-metals should cannot be completely considered explainin treating the theVCMA total effect. PMA In energy addition in theto the system. magnetic Moreover, moments both in 3 thed metals, orbital those magnetic in 5d metals moments should and bethe considered electric quadrupole in treating mechanisms the total PMA (second energy term in ofthe Equation system. (2)Moreover, and Figure both6b) the of orbital Pt dominate magnetic the momentsVCMA in and the casethe electric of L10-FePt, quadrupole as shown mechanisms in Ref. 116. (second As discussed term of Equation in the recent (2) and review Figure paper 6b) [of111 Pt], it has been widely recognized that the XAS/XMCD spectroscopy is a powerful tool to investigate the dominate the VCMA in the case of L10-FePt, as shown in Ref. 116. As discussed in the recent review papervoltage-induced [111], it has eff beenects inwidely spintronic recognized devices that [28, 113the ,116XAS/XMCD,119–124]. spectroscopy is a powerful tool to investigateA much the larger voltage-induced VCMA coeffi effectscient canin spintronic be obtained devices when [28,113,116,119–124]. compared with that of purely electronic originA ifmuch we use larger a chemical VCMA reaction coefficient [122 can,125 be]. Forobtained example, when a VCMA compared coeffi withcient that exceeding of purely 10,000 electronic fJ/Vm originoriginating if we fromuse a reversible chemical oxygenreaction ion[122,125]. migration For hasexample, been demonstrateda VCMA coefficient in the Coexceeding/GdOx system. 10,000 In Ref. 122, XAS/XMCD spectroscopy at the Co absorption edge was employed to a Ta (4 nm)/Pt fJ/Vm originating from reversible oxygen ion migration has been demonstrated in the Co/GdOx system.(3 nm)/Co In (0.9Ref. nm) 122,/GdO XAS/XMCDx (33 nm) /spectroscopyTa (2 nm)/Au at (12 th nm)e Co multilayer absorption and edge found was that employed an applied to voltagea Ta (4 changes the oxidation state and magnetization of the Co. Ref. 125 also reports real-time measurements nm)/Pt (3 nm)/Co (0.9 nm)/GdOx (33 nm)/Ta (2 nm)/Au (12 nm) multilayer and found that an applied voltageof such anchanges electrochemical the oxidation VCMA state effect. and The magnetization operating speed of the strongly Co. Ref. depends 125 also on the reports applied real-time voltage measurementsand temperature, of whichsuch an strongly electrochemical indicates that VCMA the electrochemical effect. The operating VCMA requiresspeed strongly a thermal depends activation on theprocess. applied The voltage reported and maximum temperature, speed was which in the strongly sub-millisecond indicates range.that the Therefore, electrochemical such large VCMA values requiresof the electrochemical a thermal activation VCMA seemprocess. attractive, The repo butrted lie beyond maximum the scopespeed of was VCMA in the studies sub-millisecond for working range.memory Therefore, applications. such Alarge similarly values large of the VCMA electr eochemicalffect with VCMA limited operatingseem attractive, speed but has lie been beyond observed the scopein many of systemsVCMA withstudies electrochemical for working reactionsmemory [applications.28,126,127] and A/ orsimilarly charge trapslarge [VCMA128,129 ].effect with limited operating speed has been observed in many systems with electrochemical reactions [28,126,127] and/or charge traps [128,129]. Recently, strain-induced modulation of electronic structures and its influence on the VCMA effect has attracted attention [130,131]. For example, Hibino et al. reported a high VCMA coefficient of +1600 fJ/Vm in a Pt/Co/Pd/MgO structure at 10 K [95]. Here, the thin Pd layer possesses a magnetic moment that is induced by the proximity effect from the adjacent Co layer. At room temperature, a conventional linear VCMA effect with an efficiency of −90 fJ/Vm was observed. On the other hand,

Micromachines 2019, 10, 327 12 of 31

Recently, strain-induced modulation of electronic structures and its influence on the VCMA effect has attracted attention [130,131]. For example, Hibino et al. reported a high VCMA coefficient of +1600 fJ/Vm in a Pt/Co/Pd/MgO structure at 10 K [95]. Here, the thin Pd layer possesses a magnetic moment that is induced by the proximity effect from the adjacent Co layer. At room temperature, a conventional linear VCMA effect with an efficiency of 90 fJ/Vm was observed. On the other hand, − at lower temperatures below 100 K, a strong nonlinear VCMA effect appeared with the sign reversal. They explained that the observed effect can be attributed to the temperature dependence of the strain in the Pd. Similarly, Kato et al. reported a VCMA coefficient of over +1000 fJ/Vm at room temperature in an Ir/tetragonal FeCo/MgO structure [132]. So far, only static measurements have been done in these experiments. A demonstration of a high speed response is required to confirm whether they actually originate from the purely-electronic VCMA effect or not.

4. Materials Research for a Large VCMA Effect The VCMA coefficient is one of the most important parameters for the scalability design of voltage-torque MRAM. When the cell size is reduced, we need to increase the PMA of the free layer to maintain the target thermal stability. As described in Section2, voltage-induced dynamic switching requires the elimination of the PMA during the precessional dynamics. Figure 10 shows a simple estimate of the PMA and VCMA coefficient required to consider the scalability [34,35]. As the simplest example, if we assume a free layer whose PMA is only determined by the interface magnetic anisotropy at the interface with the dielectric layer, the effective PMA energy is expressed as K (E) 1 ( ) i 2 KPMA E = µ0MS (4) tfree − 2

Here, tfree and MS are the thickness and saturation magnetization of the free layer. Ki(E) is the PMA under application of the electric-field (E), and it is given by

K (E) = K (E = 0) ηE (5) i i − where η is the VCMA coefficient. The thermal stability ∆(E) of the free layer under the application of the electric-field can be expressed by

KPMA(E)Atfree ηA ∆(E) = = ∆0 E (6) kBT − kBT

Here, A and ∆0 are the area of the free layer and the thermal stability under zero electric-field, respectively. Consequently, the VCMA coefficient, η, which is required to eliminate ∆0 can be expressed as,

k T∆ η = B 0 (7) AESW where ESW is the amplitude of the switching electric-field. For the curves in Figure 10, it was assumed that tfree = 1 nm and ESW = 1 V/nm for each value of ∆0. If we take cache memory applications as an example, the required KPMAtfree values range 2 2 from 0.2 mJ/m to 0.5 mJ/m , depending on the target ∆0 values; consequently, the required VCMA coefficient is estimated to be from 200 fJ/Vm to 500 fJ/Vm. The main memory applications need higher 2 2 KPMAtfree values in the range from 0.6 mJ/m to 1.5 mJ/m . As a result, the required VCMA coefficient is in the range from 600 fJ/Vm to 1500 fJ/Vm. However, in experiments that have only focused on the purely-electronic VCMA effect, the achieved VCMA coefficient that is demonstrated in practical materials, such as CoFeB, has been limited to about 100 fJ/Vm [71,78,81,98]. Micromachines 2019, 10, x 12 of 31 at lower temperatures below 100 K, a strong nonlinear VCMA effect appeared with the sign reversal. They explained that the observed effect can be attributed to the temperature dependence of the strain in the Pd. Similarly, Kato et al. reported a VCMA coefficient of over +1000 fJ/Vm at room temperature in an Ir/tetragonal FeCo/MgO structure [132]. So far, only static measurements have been done in these experiments. A demonstration of a high speed response is required to confirm whether they actually originate from the purely-electronic VCMA effect or not.

4. Materials Research for a Large VCMA Effect The VCMA coefficient is one of the most important parameters for the scalability design of voltage-torque MRAM. When the cell size is reduced, we need to increase the PMA of the free layer to maintain the target thermal stability. As described in Section Ⅱ , voltage-induced dynamic switching requires the elimination of the PMA during the precessional dynamics. Figure 10 shows a simple estimate of the PMA and VCMA coefficient required to consider the scalability [34,35]. As the simplest example, if we assume a free layer whose PMA is only determined by the interface magnetic anisotropy at the interface with the dielectric layer, the effective PMA energy is expressed as

𝐾(𝐸) 1 𝐾(𝐸) = − 𝜇𝑀 (4) 𝑡 2

Here, tfree and MS are the thickness and saturation magnetization of the free layer. Ki(E) is the PMA under application of the electric-field (E), and it is given by

𝐾(𝐸) =𝐾(𝐸=0) −𝜂𝐸 (5) where η is the VCMA coefficient. The thermal stability Δ(Ε) of the free layer under the application of the electric-field can be expressed by

𝐾(𝐸)𝐴𝑡 𝜂𝐴 𝛥(𝐸) = =∆ − 𝐸 (6) 𝑘𝑇 𝑘𝑇

Here, A and Δ0 are the area of the free layer and the thermal stability under zero electric-field, respectively. Consequently, the VCMA coefficient, η, which is required to eliminate Δ0 can be expressed as,

𝑘𝑇𝛥 𝜂= (7) 𝐴𝐸 Micromachines 2019, 10, 327 13 of 31 where ESW is the amplitude of the switching electric-field.

Figure 10. Scalability issue forfor voltage-torquevoltage-torque MRAMs.MRAMs. TheThe dependence dependence of of the the required requiredK PMAKPMAttfreefree and

VCMA coecoefficientfficient onon thethe diameterdiameter ofof thethe MTJMTJ waswas estimatedestimated forfor eacheach thermalthermal stabilitystability factorfactor ((∆Δ00).

We employed a fully epitaxial Cr/ultrathin Fe/MgO system as a standard system for the materials research of VCMA effect [133], because large interface magnetic anisotropy can be obtained due to the flat and well-defined Fe/MgO interface [134–136] when compared to MTJs with noble metal buffers, which can have the problem of surface segregation [137]. To evaluate the VCMA properties, we used molecular beam epitaxy to prepare orthogonally-magnetized MTJ structures that consisted of a MgO seed (3 nm)/Cr buffer (30 nm)/ultrathin Fe (tFe)/MgO (tMgO = 2.3 nm)/Fe(10 nm) on MgO(001) substrates. Here, the bottom ultrathin Fe layer is the voltage-controlled free layer with perpendicular magnetic easy axis and the top 10 nm-thick Fe is the in-plane magnetized reference layer. Figure 11a shows an example of the applied bias voltage, Vbias, and dependence of the half-MR loop measured under an in-plane magnetic field, Hex. The vertical axis is normalized using the maximum (Hex = 0 Oe) and minimum (Hex = 20 kOe) resistances. The Fe thickness is fixed at t = 0.44 nm. − Fe The application of an in-plane magnetic field tilts the magnetization of the ultrathin Fe layer into the magnetic hard axis, while that of the reference layer remains in the film plane (see the drawings in Figure 11a). Therefore, the effective perpendicular anisotropy field is reflected in the saturation behavior of tunneling resistance. The tunneling conductance, G, depends on the relative angle (θ) between the magnetizations of the free and reference layers, i.e. G(θ) = G + (G G )cosθ. Here, G 90 P− 90 90 and GP are the conductance under the orthogonal and parallel magnetization configurations. Therefore, the ratio of the in-plane component of the magnetization of the free layer, Min-plane, to its saturation magnetization, MS, is expressed as

Min plane R90 R(θ) R − = cos θ = − P (8) M R(θ) R R S 90 − P where RP is the MTJ resistance in the parallel magnetization configuration, R90 is the MTJ resistance in the orthogonal magnetization configuration, and R(θ) is the MTJ resistance when the magnetization of the ultrathin Fe layer is tilted towards the in-plane direction at angle θ under the application of an in-plane magnetic field. Using Equation (8), we can evaluate the normalized in-plane magnetization versus the applied magnetic field. The inset in Figure 11b shows an example of a normalized M-H curve measured under Vbias = 10 mV. The PMA energy, KPMA can be calculated from Min-plane (H) with the saturation magnetization value evaluated by SQUID measurements (yellow area in the inset of Figure 11b). Figure 11b summarizes the applied electric-field, Vbias/tMgO, dependence of KPMAtFe. With ultrathin layers of Fe, an unexpected nonlinear VCMA effect was observed. Under the application of negative voltages, the PMA monotonically increases with a large VCMA coefficient of 290 fJ/Vm. − On the other hand, the PMA deviates from a linear relationship under the application of positive voltages. Figure 12 summarizes the Fe thickness dependence of the VCMA coefficient. This nonlinear Micromachines 2019, 10, 327 14 of 31

VCMA effect was only observed with ultrathin layers of Fe, tFe < 0.6 nm (blue dots), and the usual linear VCMA effect appears for thicker layers (red dots). Xiang et al. systematically investigated the tunneling conductance, the PMA, and the VCMA effect in a similar system to determine the origin of the nonlinear VCMA effect, but the MgO was replaced by a MgAl2O4 barrier, which has smaller lattice mismatch with Fe. Interestingly, they found strong correlation between the VCMA effect and the quantum well states of ∆1 band formed in an ultrathin Fe layer that is sandwiched between the Cr and MgO layers [138]. These results may indicate that artificial control of the electronic states in an ultrathinMicromachines ferromagnetic 20192019,, 1010,, xx layer may provide a new approach for designing the VCMA properties.1414 ofof 3131 In addition to the influence of quantum well states, we found that intentional Cr doping at the Fe/MgO interfaceCr doping can enhance at the Fe/MgO the PMA interface and the can VCMA enhance effect the [ 62PMA]. Therefore, and the VCMA intermixing effect [62]. with Therefore, the bottom Cr intermixing with the bottom Cr buffer may also have an influence on the observed large VCMA effect. bufferintermixing may also with have the an bottom influence Cr buffer on the may observed also have large an influence VCMA on eff theect. observed A theoretical large VCMA investigation effect. to A theoretical investigation to understand the role of the inter-diffused Cr atoms has been proceeded understand the role of the inter-diffused Cr atoms has been proceeded [139,140]. [139,[139, 140].140].

FigureFigure 11. ( a11.) Bias ((aa)) voltage BiasBias voltagevoltage dependence dependencedependence of normalized ofof normalizednormalized tunnel tunneltunnel magnetoresistance magnetoresistancemagnetoresistance (TMR) (TMR)(TMR) curves curvescurves measured undermeasured in-plane under magnetic in-plane fields magnetic for an orthogonallyfields for an magnetizedorthogonally MTJmagnetized consisting MTJ ofconsisting Cr/ultrathin of Fe (0.44Cr/ultrathin nm)/MgO/ FeFe (0.44 (10 nm). nm)/MgO/Fe The inset (10 showsnm). The a in cross-sectionalsetset showsshows aa cross-sectionalcross-sectional TEM image TEMTEM of the imageimage MTJ. ofof thethe (b ) MTJ.MTJ. Applied ((b)) AppliedApplied electric-fieldelectric-field dependencedependence ofof KPMAPMAttFeFe values.values. TheThe insetinset displaysdisplays anan exampleexample ofof aa normalizednormalized electric-field dependence of KPMAtFe values. The inset displays an example of a normalized M-H curve. M-H curve. KPMA was evaluated from the yellow-colored area with the saturation magnetization value K Mwas-H curve. evaluated KPMA was from evaluated the yellow-colored from the yellow-colored area with area the with saturation the saturation magnetization magnetization value value that was PMAthat was obtained by a SQUID measurement. Reprinted figure with permission from [133], Copyright obtainedthat was by aobtained SQUID by measurement. a SQUID measurement. Reprinted Reprin figureted with figure permission with permissi fromon from [133], [133], Copyright Copyright 2017 by 20172017 byby thethe AmericanAmerican Physical Society. the American Physical Society.

FigureFigure 12. Fe 12. thickness Fe thickness dependence dependence of the VCMAof the coeVCMAfficient co observedefficientefficient observed inobserved a Cr/ultrathin inin aa Cr/ultrathinCr/ultrathin Fe(tFe)/MgO /Fe

structure.Fe(ttFeFe)/MgO/Fe)/MgO/Fe A large VCMA structure.structure. coe AA ffi largelargecient VCMAVCMA with coefficientcoefficient nonlinear wiwi behaviorthth nonlinearnonlinear was behaviorbehavior found was inwas the foundfound thinner inin thethe Fe thinnerthinner thickness Fe thickness range, tFe < 0.6 nm (blue dots). Reprinted figure with permission from [133], Copyright range,Fet Fethickness< 0.6 nm range, (blue tFe dots). < 0.6 Reprintednm (blue dots). figure Reprinted with permission figure with from permission [133], Copyright from [133], by Copyright the American Physicalby the Society. American Physical Society.

A large VCMA effect can be obtained with the Cr/ultrathin Fe/MgO system; however, we can only induce an enhancement in the PMA. As explained in section Ⅱ,, reductionreduction inin thethe PMAPMA isis requiredrequired forfor voltage-inducedvoltage-induced dynamicdynamic switchingswitching ofof thethe perpendicularly-magnetizedperpendicularly-magnetized freefree layer.layer. Nakamura et al. proposed inserting a heavy metal monolayerlayer atat thethe Fe/MgOFe/MgO interfaceinterface toto improveimprove thethe VCMAVCMA properties,properties, andand foundfound usingusing first-principlesfirst-principles calculationscalculations thatthat 55d transitiontransition metals,metals, suchsuch asas IrIr andand Os,Os, wouldwould bebe effectiveeffective inin enhancinenhancing the VCMA coefficient [141]. A few experimental

Micromachines 2019, 10, 327 15 of 31

A large VCMA effect can be obtained with the Cr/ultrathin Fe/MgO system; however, we can only induce an enhancement in the PMA. As explained in Section2, reduction in the PMA is required for voltage-induced dynamic switching of the perpendicularly-magnetized free layer. Nakamura et al. proposed inserting a heavy metal monolayer at the Fe/MgO interface to improve the VCMA properties, and found using first-principles calculations that 5d transition metals, such as Ir and Os, would be effective in enhancing the VCMA coefficient [141]. A few experimental trials of interface engineering that included the insertion of a heavy metal layer at a CoFe-based film/MgO interfaceMicromachines have been2019, 10 reported, x [81,142]; however, the VCMA coefficient was still less than15 100 of 31 fJ /Vm. Ir seemstrials to of be interface a promising engineering candidate that included for this the purpose insertion due of to a itsheavy huge metal spin-orbit layer at coupling a CoFe-based constant, whichfilm/MgO is more interface than 10 timeshave been larger repo thanrted that [81,142]; of 3d however,transition the ferromagnets VCMA coefficient [141]. was still less than We100 fJ/Vm. prepared Ir seems multilayer to be a structurespromising candidate consisting for of this Cr purpose (30 nm) /dueultrathin to its huge Fe(t Fespin-orbit)/Ir(tIr)/MgO coupling (2.5 nm) withconstant, indium-tin which oxide is more (ITO) than or Fe10 times (10 nm) larger top than electrodes that of 3 tod transition investigate ferromagnets the impact [141]. of the introduction of Ir on theWe interfacial prepared multilayer PMA and structures the VCMA cons effistingect [35 of]. Cr The (30 ultrathinnm)/ultrathin Ir layer Fe(tFe was)/Ir(t insertedIr)/MgO (2.5 between nm) the Fe andwith MgO indium-tin layers; oxide however, (ITO) we or foundFe (10 nm) that top the electrodes Ir atoms to were investigate dispersed the insideimpact theof the Fe introduction layer during the post-annealingof Ir on the process,interfacial as PMA seen and in thethe HAADF-STEMVCMA effect [35]. images The ultrathin in Figure Ir layer 13a. was Atomic-scale inserted betweenZ-contrast the Fe and MgO layers; however, we found that the Ir atoms were dispersed inside the Fe layer during HAADF-STEM imaging enabled the identification of inter-diffused Ir atoms as bright spots that are the post-annealing process, as seen in the HAADF-STEM images in Figure 13a. Atomic-scale Z- indicated by yellow arrows. The first-principles calculation predicts strong in-plane anisotropy at the contrast HAADF-STEM imaging enabled the identification of inter-diffused Ir atoms as bright spots Ir/MgOthatinterface are indicated [141]; by however, yellow wearrows. observed The first-principles an unexpected calculation enhancement predicts in the strong PMA. in-plane Figure 13b showsanisotropy a comparison at the Ir/MgO between interface the polar [141]; MOKE however, hysteresis we observed curves an unexpected of a single enhancement Fe layer (tFe in= the1.0 nm) and anPMA. Ir-doped Figure Fe13b layer shows formed a comparison the bilayer between structure the pola consistingr MOKE hysteresis of Fe (1.0 curves nm) /ofIr a (0.1 single nm)). Fe layer The pure Fe layer(tFe = exhibits 1.0 nm) and large an saturation Ir-doped Fe fields layer offormed about the 7 bilayer kOe, which structure indicated consisting an of in-plane Fe (1.0 nm)/Ir magnetic (0.1 easy axis.nm)). On the The other pure Fe hand, layer the exhibits introduction large saturation of the quitefields of thin about Ir doping7 kOe, which layer indicated resulted an in in-plane transition of the magneticmagnetic easy axis. axis On from the theother in-plane hand, the to introduction the out-of-plane of the quite direction. thin Ir Figuredoping 13layerc summarizes resulted in the transition of the magnetic easy axis from the in-plane to the out-of-plane direction. Figure 13c dependence of the intrinsic interfacial magnetic anisotropy, Ki,0, on the thickness of the Ir layer. With summarizes the dependence of the intrinsic2 interfacial magnetic anisotropy, Ki,0, on the thickness of appropriate Ir doping, Ki,0 reaches 3.7 mJ/m , which is about 1.8 times that observed at the Fe/MgO the Ir layer. With appropriate Ir doping, Ki,0 reaches 3.7 mJ/m2, which is about 1.8 times that observed interface (2.0 mJ/m2)[35,134]. at the Fe/MgO interface (2.0 mJ/m2) [35,134].

FigureFigure 13. 13.(a )(a HAADF-STEM) HAADF-STEM images images of ofa multilayer a multilayer structure structure of Cr/ultrathin of Cr/ultrathin Ir-doped Fe/MgO. Ir-doped Inter- Fe/MgO. Inter-didiffusedffused Ir Ir atoms atoms can can be beidentified identified by atomic-scale by atomic-scale Z-contrast Z-contrast HAADF-STEM HAADF-STEM imaging imaging as indicated as indicated by by thethe yellow yellow arrows.arrows. ( (bb) )Comparison Comparison of the of thepolar polar MOKE MOKE hysteresis hysteresis curves curves for pure for Fe pure(1 nm)/MgO Fe (1 nm) and/MgO and FeFe (1(1 nm)nm)/Ir/Ir (0.1 nm)/MgO nm)/MgO structures. structures. (c () cDependence) Dependence of the of the intrinsic intrinsic interface interface magnetic magnetic anisotropy anisotropy energy, Ki,0, on the thickness of the Ir layer. Reproduced from [35]. CC BY 4.0. energy, Ki,0, on the thickness of the Ir layer. Reproduced from [35]. CC BY 4.0. The Ir doping also has an effect on the VCMA. Figure 14 a shows an example of the bias voltage effect on the TMR curves that were measured under in-plane magnetic fields for an orthogonally- magnetized MTJ with an Ir-doped Fe free layer (tFeIr = 0.82 nm; formed from Fe (0.77 nm)/Ir (0.05 nm)). The saturation field shifts with changes in the applied voltage, as is the case in a pure Fe/MgO structure. However, the applied electric-field dependence of KPMAtFeIr exhibits a completely different

Micromachines 2019, 10, 327 16 of 31

The Ir doping also has an effect on the VCMA. Figure 14a shows an example of the bias voltage effect on the TMR curves that were measured under in-plane magnetic fields for an orthogonally-magnetized MTJ with an Ir-doped Fe free layer (tFeIr = 0.82 nm; formed from Fe (0.77 nm)/Ir (0.05 nm)). The saturation field shifts with changes in the applied voltage, as is the case in a pure Fe/MgO structure. However, the Micromachines 2019, 10, x 16 of 31 applied electric-field dependence of KPMAtFeIr exhibits a completely different trend when compared withtrend that when observed compared in thewith Fe that/MgO observed structure. in the We Fe observed/MgO structure. a large reductionWe observed in PMA a large with reduction a VCMA in coefficient of 320 fJ/Vm under positive voltages (see Figure 14b). It is interesting that such a low PMA with a VCMA− coefficient of −320 fJ/Vm under positive voltages (see Figure 14b). It is interesting dopingthat such concentration a low doping of concentration Ir, which is even of Ir, thinner which thanis even one thinner monolayer, than canone havemonolayer, a drastic can eff haveect on a thedrastic VCMA effect properties. on the VCMA In addition, properties. voltage-induced In addition, voltage-induced FMR measurements FMR confirmedmeasurements the high confirmed speed responsethe high speed of the response VCMA e offfect, the asVCMA shown effect, in the as inset shown in Figurein the inset 14b. in Thus, Figure the 14b. observed Thus, largethe observed VCMA comeslarge VCMA from purely-electronic comes from purely-electronic origin. origin.

Figure 14. ((a)) Bias Bias voltage voltage dependence dependence of of normalized normalized TM TMRR curves measured under in-plane magnetic fieldsfields for an orthogonally-magnetized MTJ consistingconsisting of Cr/Ir-doped Cr/Ir-doped Fe(0.82 nm)/MgO/Fe(10 nm)/MgO/Fe(10 nm). ( b)

Applied electric-fieldelectric-field dependencedependence ofofK KPMAPMAtFeIr. .The The inset inset shows shows an an example example of of voltage-induced FMRFMR excitation measuredmeasured by by a homodyne a homodyne detection detectio technique,n technique, which proves which the highproves speed the responsiveness high speed ofresponsiveness the observed of VCMA the observed effect. Reproduced VCMA effect. from Reproduced [35]. CC BY from 4.0. [35]. CC BY 4.0.

A theoreticaltheoretical analysisanalysis using using first-principles first-principles calculation calculation was performedwas performed in Cu(5ML) in/ FeCu(5ML)/Fe94Ir6(5ML)94/MgO(5ML)Ir6(5ML)/MgO(5ML) structures structures to discuss to the discuss physical the origin physical of the origin large of VCMA the large effect VCMA in Ir-doped effect Fe.in Ir-doped The Ir-doped Fe. The bcc Ir-doped Fe was modeled bcc Fe was by amodeled supercell by consisting a supercell of 4consisting4 unit cells of 4×4 as shown unit cells in Figure as shown 15a. × Figurein Figure 15b 15a. depicts Figure the atomic-resolved15b depicts the electric-fieldatomic-resolved induced electric-field magnetic anisotropyinduced magnetic energies anisotropy (MAE) for theenergies Fe and (MAE) Ir atoms. for the The Fe variation and Ir atoms. in the The MAE variation for the in Ir atomsthe MAE is morefor the than Ir atoms five times is more greater than than five thattimes for greater the Fe than atoms. that Interestingly, for the Fe atoms. MAE changeInterestin ingly, the secondMAE change layer (layerin the 2 second in Figure layer 15b) (layer from 2 the in interfaceFigure 15b) with from the MgOthe interface layer is largerwith the than MgO that oflayer the layeris larger 1, contrary than that to expectations.of the layer 1, contrary to expectations.We also attempted to divide the MAE into contributions from the spin-flip and spin-conserved termsWe between also attempted the occupied to divide and unoccupiedthe MAE into states. contri Figurebutions 15 fromc shows the thespin-flip voltage-induced and spin-conserved changes interms MAE between that arise the from occupied second-order and unoccupied perturbation states of. Figure the Ir sites15c shows in layers the 1voltage-induced and 2. The electric-field changes modulationin MAE that of arise the spin-conservedfrom second-order term perturbation for the majority of the spin Ir sites occupied in layers and 1 unoccupied and 2. The stateselectric-fieldδE is ↑↑ largermodulation than that of the for spin-conserved the minority spin term states for δtheE majority. On the otherspin occupied hand, the and spin-flip unoccupied terms that states are δ byE↑↑ the is ↓↓ electric-field,larger than thatδE forand theδ Eminorityhave almostspin states the same δE↓↓ absolute. On the value,other hand, but with the opposite spin-flip sign, terms so that the VCMAare by ↑↓ ↓↑ etheffect electric-field, that arises fromδE↑↓ theand spin-flip δE↓↑ have term almost is small. the same Therefore, absolute the valu largee, VCMA but with eff oppositeect in Ir-doped sign, so Fe the is mainlyVCMA causedeffect that by arises the electric-field from the spin-flip effect on term the is majority small. Therefore, spin Ir-5d thestates large and VCMA it can beeffect interpreted in Ir-doped by theFe is modulation mainly caused in the by firstthe electric-field term of Equation effect (2), on i.e.the majority the orbital spin magnetic Ir-5d states moment and it mechanism. can be interpreted by the modulation in the first term of Equation (2), i.e. the orbital magnetic moment mechanism.

Micromachines 2019, 10, 327 17 of 31 Micromachines 2019, 10, x 17 of 31

FigureFigure 15. 15.First First principles principles calculations calculations of the of electric-fieldthe electric-field induced induced magnetic magnetic anisotropy anisotropy energy energy change in anchange Ir-doped in an FeIr-doped/MgO system.Fe/MgO (system.a) Supercell (a) Supercell structure structure used for used the for calculation, the calculation, consisting consisting of MgO of (5 ML)MgO/FeIr (5 ML)/FeIr (5 ML)/MgO (5 ML)/MgO (5 ML). ((5b )ML). Atomic-resolved (b) Atomic-resolved magnetic magnetic anisotropy anisotropy energies energies (MAE) (MAE) change inducedchange by induced an electric-field by an electric-field of 0.1 V of/nm 0.1 inV/nm MgO. in MgO. The Ir The concentration Ir concentration was was maintained maintained at at about about 6% in the6% in FeIr the layer. FeIr layer. (c) The (c) The electric-field electric-field induced induced MAE MAE arising arising from from second-ordersecond-order perturbation perturbation of of the the spin-orbitspin-orbit coupling coupling for for Ir Ir atoms atoms in in layers layers 1 1 and and 2. 2. ReproducedReproduced from [35]. [35]. CC CC BY BY 4.0. 4.0.

FigureFigure 16 16shows shows the the density density of of states states forfor IrIr atomsatoms inin layerlayer 2. The The majority majority spin spin 5d 5 dstatesstates are are dominantdominant near near the the Fermi Fermi level, level, since since the the minority minority spinspin 5d states near near the the Fermi Fermi level level form form bonding bonding andand anti-bonding anti-bonding states states by by hybridizationhybridization with with the the minority minority spin spin Fe-3 Fe-3d states.d states. On the On other the otherhand, hand,the themajority majority spin spin 5d 5 statesd states are are well-localized well-localized when when compared compared with the with minority the minority spin states spin near states the nearFermi the Fermilevel. level. Figure Figure 16 also 16 also shows shows the theMAE MAE as a as functi a functionon of ofthe the Fermi Fermi energy energy shift shift (black (black line). line). The The PMA PMA energy is drastically modified by a small shift in Fermi energy reflecting the localized majority spin energy is drastically modified by a small shift in Fermi energy reflecting the localized majority spin states and the large spin-orbit coupling of the Ir atoms. As a result, a large VCMA is obtained for the states and the large spin-orbit coupling of the Ir atoms. As a result, a large VCMA is obtained for the charge-doping effect even in layer 2. charge-doping effect even in layer 2. MicromachinesThe theoretical 2019, 10, x calculations predict the larger VCMA effect exceeding a few thousand fJ/Vm18 of 31by inserting a monolayer of Ir at the Fe/MgO interface; however, such a structure can drastically degrade the TMR properties in the MTJ device, in addition to the strong in-plane anisotropy. On the other hand, if Ir doping can improve both the PMA and the VCMA effect while minimizing degradation in TMR, the MTJs should be much more manufacturable, even by sputtering processes. In fact, the enhancement of the PMA and the VCMA effect by Ir doping has also been confirmed in polycrystalline MTJs that are mainly prepared by sputtering [143]. We still have numerous choices for the 4d and 5d elements, therefore materials engineering using heavy metal doping has enormous possibilities for further improvement in the interfacial PMA and VCMA properties.

FigureFigure 16. 16.Spin Spin polarized polarized locallocal densitydensity of states of of Ir-5 Ir-5dd orbitalsorbitals and and magnetic magnetic anisotropy anisotropy energy energy as a as a functionfunction of of the the band band energy energy in in layer layer 2.2.

5. Towards Reliable Voltage-Induced Dynamic Switching In this section, recent experimental trials for reliable voltage-induced dynamic switching are discussed. As shown in Figure 4, voltage-driven magnetization switching is initiated by precession of the magnetization that is induced by the VCMA effect and the associated voltage-torque, which is proportional to the time derivative of the applied voltage. During the application of a voltage, the magnetization precesses around the effective field while undergoing magnetization damping. Once the voltage is turned off, the magnetic anisotropy immediately recovers as the ferromagnetic layer/dielectric layer junction discharges, and the magnetization relaxes into one of two polarities. We can achieve bipolar magnetization switching using a unipolar voltage pulse with a controlled duration since the polarity of the final state can be controlled by the voltage pulse width. In the absence of thermal fluctuations, the magnetization trajectory during the switching process is uniquely determined for a given initial state and voltage pulse shape, and therefore error-free magnetization switching can be achieved by choosing the appropriate voltage pulse width. However, in practice, the magnetization inevitably suffers thermal fluctuations and that results in the stochastic generation of write errors. Special care must be taken when attempting to reduce the write errors in voltage-torque MRAM cells. In the case of STT, the current polarity determines the polarity of magnetization switching, and a longer pulse may be used to reduce write errors. On the other hand, in the case of voltage-induced dynamic switching, a longer pulse dampens the magnetization along the effective field direction, and this degrades the switching accuracy. Although earlier experiments have characterized the basics of voltage-driven magnetization switching, it was only in 2016 that the WER in a practical MTJ was quantitatively evaluated for the first time [105]. Figure 17a shows a schematic illustration of an experimental setup for evaluating the WER of an MTJ. Voltage pulses that were generated by the pulse generator are fed to the MTJ and these switch the free layer magnetization. The free layer magnetization direction, either parallel or antiparallel with respect to the reference layer magnetization, can be monitored via the TMR effect. Figure 17b displays the typical behavior of voltage-driven magnetization switching; Psw is the switching probability, tpulse is the pulse width; and, Vpulse is the voltage amplitude. When Vpulse is small, the VCMA effect cannot completely eliminate the magnetic energy barrier; therefore, the magnetization switching in this region is dominated by thermal activation. As Vpulse is increased, well- defined oscillation of Psw appears, which is a signature of precession-mediated switching induced by the VCMA effect. As discussed in Section Ⅱ, the highest Psw is obtained at tpulse that corresponds to one-half the magnetization precession cycle, and then Psw gradually moves toward 0.5 while

Micromachines 2019, 10, 327 18 of 31

The theoretical calculations predict the larger VCMA effect exceeding a few thousand fJ/Vm by inserting a monolayer of Ir at the Fe/MgO interface; however, such a structure can drastically degrade the TMR properties in the MTJ device, in addition to the strong in-plane anisotropy. On the other hand, if Ir doping can improve both the PMA and the VCMA effect while minimizing degradation in TMR, the MTJs should be much more manufacturable, even by sputtering processes. In fact, the enhancement of the PMA and the VCMA effect by Ir doping has also been confirmed in polycrystalline MTJs that are mainly prepared by sputtering [143]. We still have numerous choices for the 4d and 5d elements, therefore materials engineering using heavy metal doping has enormous possibilities for further improvement in the interfacial PMA and VCMA properties.

5. Towards Reliable Voltage-Induced Dynamic Switching In this section, recent experimental trials for reliable voltage-induced dynamic switching are discussed. As shown in Figure4, voltage-driven magnetization switching is initiated by precession of the magnetization that is induced by the VCMA effect and the associated voltage-torque, which is proportional to the time derivative of the applied voltage. During the application of a voltage, the magnetization precesses around the effective field while undergoing magnetization damping. Once the voltage is turned off, the magnetic anisotropy immediately recovers as the ferromagnetic layer/dielectric layer junction discharges, and the magnetization relaxes into one of two polarities. We can achieve bipolar magnetization switching using a unipolar voltage pulse with a controlled duration since the polarity of the final state can be controlled by the voltage pulse width. In the absence of thermal fluctuations, the magnetization trajectory during the switching process is uniquely determined for a given initial state and voltage pulse shape, and therefore error-free magnetization switching can be achieved by choosing the appropriate voltage pulse width. However, in practice, the magnetization inevitably suffers thermal fluctuations and that results in the stochastic generation of write errors. Special care must be taken when attempting to reduce the write errors in voltage-torque MRAM cells. In the case of STT, the current polarity determines the polarity of magnetization switching, and a longer pulse may be used to reduce write errors. On the other hand, in the case of voltage-induced dynamic switching, a longer pulse dampens the magnetization along the effective field direction, and this degrades the switching accuracy. Although earlier experiments have characterized the basics of voltage-driven magnetization switching, it was only in 2016 that the WER in a practical MTJ was quantitatively evaluated for the first time [105]. Figure 17a shows a schematic illustration of an experimental setup for evaluating the WER of an MTJ. Voltage pulses that were generated by the pulse generator are fed to the MTJ and these switch the free layer magnetization. The free layer magnetization direction, either parallel or antiparallel with respect to the reference layer magnetization, can be monitored via the TMR effect. Figure 17b displays the typical behavior of voltage-driven magnetization switching; Psw is the switching probability, tpulse is the pulse width; and, Vpulse is the voltage amplitude. When Vpulse is small, the VCMA effect cannot completely eliminate the magnetic energy barrier; therefore, the magnetization switching in this region is dominated by thermal activation. As Vpulse is increased, well-defined oscillation of Psw appears, which is a signature of precession-mediated switching induced by the VCMA effect. As discussed in Section2, the highest Psw is obtained at tpulse that corresponds to one-half the magnetization precession cycle, and then Psw gradually moves toward 0.5 while undergoing damped oscillations. This behavior can be understood as the combined action of magnetization damping and thermal fluctuations. In Ref. 105, Shiota et al. employed perpendicularly magnetized MTJ (p-MTJ) that consisted of a reference layer/MgO/Fe B /W cap and experimentally demonstrated a WER of 4 10 3. They also 80 20 × − demonstrated in numerical simulations that the WER could be reduced by improving the thermal stability factor, ∆ and by reducing the magnetic damping, α of the free layer, as shown in Figure 17c. An improved ∆ effectively reduces the thermal fluctuations in the initial state and in the relaxation process after switching. Moreover, a lower α can reduce the influence of thermal fluctuations during Micromachines 2019, 10, x 19 of 31 undergoing damped oscillations. This behavior can be understood as the combined action of magnetization damping and thermal fluctuations. In Ref. 105, Shiota et al. employed perpendicularly magnetized MTJ (p-MTJ) that consisted of a reference layer/MgO/Fe80B20/W cap and experimentally demonstrated a WER of 4 × 10−3. They also demonstrated in numerical simulations that the WER could be reduced by improving the thermal α stabilityMicromachines factor,2019 ,Δ10 and, 327 by reducing the magnetic damping, of the free layer, as shown in Figure19 17c. of 31 An improved Δ effectively reduces the thermal fluctuations in the initial state and in the relaxation process after switching. Moreover, a lower α can reduce the influence of thermal fluctuations during the switching process, which leads to more accurate writing. However, However, it it should be noted that, the larger the value value of of Δ∆,, the the larger larger the the VCMA VCMA efficiency efficiency requir required,ed, otherwise otherwise the the magnetization magnetization switching switching is dominated dominated by by thermal thermal activation, activation, and and well-controlled well-controlled magnetization magnetization switching switching cannot cannot be be obtained. obtained. By using CoFeB/MgO/CoFeB CoFeB/MgO/CoFeB p-MTJs, Grezes et al. experimentally investigated the WER and the read disturbance raterate as as a functiona function of read of/ writeread/write pulse widthpulse and width amplitude, and amplitude, and examined and the examined compatibility the compatibilityof the bit-level of device the bit-level performance device for performance integration for with integration CMOS processes with CMOS [110]. processes They also [110]. simulated They alsothe performancesimulated the of performance a 256 kbit voltage-torque of a 256 kbit voltage-torque MRAM block MRAM in a 28 nmblock CMOS in a 28 process, nm CMOS and process, showed 9 andthe capabilityshowed the of capability the MTJs forof the delivering MTJs for WERs delivering below WERs 10− for below 10 ns 10 total−9 for write 10 ns time total by write introducing time by introducingthe read verify the processes.read verify The processes. introduction The introducti of read verifyon of read processes verify makes proces itses possible makes toit possible reduce the to reduceeffective the WER, effective however WER, it however causes an it causes increase an in incr theease total in writingthe total time. writing Therefore, time. Therefore, we need we further need furthereffort to effort reduce to reduce the essential the essential WER that WER is that induced is induced by single by single pulse pulse switching. switching. Recently, Recently, Shiota Shiota et al. etshowed al. showed that improvement that improvement in the PMAin the and PMA VCMA and propertiesVCMA properties can be achieved can be inachieved the MTJ in consisting the MTJ 5 of Ta/(Co Fe ) B /MgO/reference layer, and demonstrated a WER of 2 10 without−5 the read consisting30 of Ta/(Co70 80 2030Fe70)80B20/MgO/reference layer, and demonstrated a WER× of− 2 × 10 without the readverify verify process process [106 ].[106]. Further Further optimization optimization of of the the composition composition of of the the CoFeB CoFeB alloy alloy andand thethe device 6 structure allowed for a WER lower thanthan 1010−−6 toto be be achieved, achieved, as as shown shown in in Figure Figure 1818 [[109].109]. In In this this case, case, 12 the introduction of a once readread verify process enables a practicalpractical WERWER ofof thethe orderorder ofof 1010−-12. .

Figure 17. ((aa)) Experimental Experimental setup setup for for evalua evaluatingting the the WER WER of of an MTJ. (b) Pulsed-voltage-drivenPulsed-voltage-driven magnetization switching in a p-MTJ. ( c) WER as as a a function function of of Δ∆ obtainedobtained from from numerical numerical simulations. simulations.

In additionaddition to materialsto materials engineering, engineering, a physical a ph understandingysical understanding of the voltage-driven of the voltage-driven magnetization magnetizationdynamics is also dynamics needed inis orderalso needed to facilitate in order reductions to facilitate in the reductions WER. Recent in the studies WER. [107 Recent,108] showedstudies [107,108]that numerical showed simulations that numerical that aresimulations based on that the are macrospin based on approximation the macrospin couldapproximation well reproduce could wellthe experimentalreproduce the data experimental by taking data into by account taking in thermalto account fluctuations thermal fluctuations and magnetization and magnetization damping. damping.In macrospin In macrospin approximation, approximation, the free layer the spinsfree layer are representedspins are represented by a magnetic by a momentmagneticM moment, and its Mtime, and evolution its time can evolution be obtained can by be numerically obtained by solving numerically the Landau-Lifshitz-Gilbert solving the Landau-Lifshitz-Gilbert equation: equation: dM αM dM = γM Heff + (9) dt × Ms × dt where Ms is the saturation magnetization, t is the time, α is the damping constant, and Heff is the effective field given by dE H = (10) eff −dM Micromachines 2019, 10, x 20 of 31

𝑑𝑴 𝛼𝑴 𝑑𝑴 = 𝛾𝑴×𝑯 + × (9) 𝑑𝑡 𝑀 𝑑𝑡 where Ms is the saturation magnetization, t is the time, α is the damping constant, and Heff is the Micromachineseffective field2019 given, 10, 327 by 20 of 31 𝑑𝐸 𝑯 = − (10) and E is the energy density expressed as 𝑑𝑴 and E is the energy density expressed as  2 E = K 1 m MsHxmx (11) PMA − z − 𝐸= 𝐾 (1 − 𝑚)−𝑀𝐻𝑚 (11) where m = (mx, my, mz) is the magnetization unit vector and Hx is an in-plane bias magnetic field. Aswhere displayed m = (m inx, m Figurey, mz) is19 thea, without magnetization the VCMA unit e ffvectorect, the and magnetization Hx is an in-plane hastwo bias energy magnetic equilibrium field. As  q  displayed in Figure 19a, without2 the VCMA effect, the magnetization has two energy equilibrium at at me = mex, 0, 1 mex , where mex = MsHx/(2KPMA), one maximum at mx = 1, and one saddle ± ± − − 𝒎 ± =(𝑚,0,±1−𝑚), where 𝑚 =𝑀𝐻/(2𝐾), one maximum at mx = −1, and one saddle point point at mx = 1. By letting KPMA fall to zero, the magnetization precesses around Hx associated with at mx = 1. By letting KPMA fall to zero, the magnetization precesses around Hx associated with damping, damping, and the appropriate duration can switch the magnetization direction. and the appropriate duration can switch the magnetization direction. Figure 19b displays a typical plot of the dependence of WER on tpulse that was observed in an Figure 19b displays a typical plot of the dependence of WER on tpulse that was observed in an MTJ consisting of a Ta/(Co30Fe70)80B20 (1.1 nm)/MgO/reference layer. The amplitude of the in-plane MTJ consisting of a Ta/(Co30Fe70)80B20 (1.1 nm)/MgO/reference layer. The amplitude of the in-plane component of the bias magnetic field is 890 Oe. The filled circles and the line denote data were component of the bias magnetic field is 890 Oe. The filled circles and the line denote data were obtained from experiments and numerical simulations, respectively [107]. Good agreement with obtained from experiments and numerical simulations, respectively [107]. Good agreement with the the experimental data suggests the validity of the model used for the numerical simulations. It is experimental data suggests the validity of the model used for the numerical simulations. It is noteworthy that the WER exhibits a local maximum at a certain tpulse, which cannot be explained noteworthy that the WER exhibits a local maximum at a certain tpulse, which cannot be explained just just by considering the VCMA effect. A detailed analysis of the magnetization trajectory revealed by considering the VCMA effect. A detailed analysis of the magnetization trajectory revealed that that thermal agitation during the relaxation process (i.e., after the pulse application) induces the thermal agitation during the relaxation process (i.e., after the pulse application) induces the transition transition of the magnetization between the precession orbits surrounding the energy minima and that of the magnetization between the precession orbits surrounding the energy minima and that the the precession-orbit transition enhances the WER. The numerical simulations also revealed that the precession-orbit transition enhances the WER. The numerical simulations also revealed that the probability of the precession-orbit transition depends on tpulse (see Ref. 107 for more details). In the probability of the precession-orbit transition depends on tpulse (see Ref. 107 for more details). In the present case, the probability is maximized at around tpulse = 0.12 ns. This results in the appearance of a present case, the probability is maximized at around tpulse = 0.12 ns. This results in the appearance of local maximum in the WER, and it narrows the operating tpulse range for which reliable magnetization a local maximum in the WER, and it narrows the operating tpulse range for which reliable switching is assured. As the appearance of the WER local maximum is related to magnetization magnetization switching is assured. As the appearance of the WER local maximum is related to fluctuations during the relaxation process, we need to reduce its influence by improving the PMA and magnetization fluctuations during the relaxation process, we need to reduce its influence by VCMA properties in order to achieve a wide operating tpulse range. improving the PMA and VCMA properties in order to achieve a wide operating tpulse range.

Figure 18. Example of the optimized WER as a function of tpulse observed in a perpendicularly- Figure 18. Example of the optimized WER as a function of tpulse observed in a perpendicularly- magnetized MTJ consisting of Ta/(Co50Fe50)80B20/MgO/reference layer. The blue and red symbols magnetized MTJ consisting of Ta/(Co50Fe50)80B20/MgO/reference layer. The blue and red symbols represent the WER of parallel (P) to antiparallel (AP) and AP to P switching, respectively. Reprinted represent the WER of parallel (P) to antiparallel (AP) and AP to P switching, respectively. Reprinted figure with permission from [109], Copyright 2019 by the IOP Publishing Ltd. figure with permission from [109], Copyright 2019 by the IOP Publishing Ltd.

Micromachines 2019, 10, 327 21 of 31 Micromachines 2019, 10, x 21 of 31

Figure 19.19. ((aa)) ContourContour plot plot of of energy energy density density in in the the absence absence of aof bias a bias voltage. voltage. (b) Appearance(b) Appearance of a localof a peaklocal inpeak the in WER the observedWER observed in an MTJin an consisting MTJ consisting of Ta/(Co of 30Ta/(CoFe70)8030BFe2070)(1.180B20 nm) (1.1/MgO nm)/MgO/reference/reference layer. Thelayer. filled The circles filled andcircles the linesand representthe lines experimentalrepresent experimental data and numerical data and simulations, numerical respectively.simulations, Reprintedrespectively. figure Reprinted with permission figure with from permission [107], Copyright from [107], 2018 Copyrigh by the Americant 2018 by the Physical American Society. Physical Society. In addition to tpulse, a recent study revealed that the WER depends in a unique manner on the rise timeIn addition (trise) and to t fallpulse time, a recent (tfall)[ study108]. revealed Figure 20 that displays the WER the magnetizationdepends in a unique trajectories manner that on were the obtainedrise time by(trise using) and threefall time different (tfall) waveforms.[108]. Figure When20 displays a pulsed the voltage magnetization is applied, trajectories the magnetization that were rotates from m+ towards m (red line) and, after the pulse, the magnetization relaxes to either me+ or obtained by using three different− waveforms. When a pulsed voltage is applied, the magnetization mrotatese , depending from m+ ontowardstpulse (greenm− (red line). line) Anand, important after the pulse, thing isthe that, magnetization due to the nonzerorelaxes to magnetization either 𝑚 or − damping,𝑚 , depending the magnetization on tpulse (green direction line). An at important the end of thing the voltage is that, pulsedue to (m the0) never nonzero reaches magnetizationme+ or me − whateverdamping, tthepulse magnetizationis chosen as long direction as one at uses the square end of pulses the voltage (Figure pulse 20a). (m Therefore,’) never reaches it takes 𝑚 some or time 𝑚 beforewhatever the t magnetizationpulse is chosen as settles long downas one to uses the square energy pulses minimum. (Figure During 20a). that Therefore, time, the it magnetizationtakes some time is subjectedbefore the tomagnetization thermal agitation, settles and down a finite to the number energy of minimum. write errors Duri willng bethat counted. time, the When magnetization a nonzero tisrise subjectedand/ornonzero to thermaltfall agitation,is introduced, and a finite the magnetization number of write is subjected errors will not be only counted. to H xWhen, but also a nonzero to the anisotropytrise and/or nonzero field due t tofall theis introduced, uncompensated the magnetizat PMA KPMAion0(V is,t ),subjected which is not given only by to Hx, but also to the anisotropy field due to the uncompensated PMA KPMA’(V,t), which is given by ( ) 2KPMA0 V, t mz Hani = (12) − 2𝐾M(𝑉,s 𝑡)𝑚 𝐻 = − (12) 𝑀 Since the polarity of Hani switches according to the polarity of mz, it applies additional torque to the magnetizationSince the polarity that tilts of the H magnetizationani switches according to Hx during to thetrise polarity(Figure of 20 mb),z, it and applies it pulls additional the magnetization torque to awaythe magnetization from Hx during thattfall tilts(Figure the magnetization20c). As a result, to forHx triseduring= 0.085 trise ns,(Figurem0 comes 20 b), closer and to it the pulls saddle the point,magnetization whereas, away for tfall from= 0.085 Hx during ns, m0 talmostfall (Figure overlaps 20c). As with a result,me and for thereby trise = 0.085 one canns, m minimize’ comes closer the time to − thatthe saddle is required point, for whereas, relaxation. for This tfall suggests= 0.085 ns, that m there’ almost is a certainoverlapstfall withwhich 𝑚 can and minimize therebythe one WER. can Indeed,minimize such the WERtime reductionthat is required is experimentally for relaxation. obtained This suggests and the that numerical there is simulations a certain tfall reproduce which can it, asminimize shown inthe Figure WER. 20 Indeed,d,e. such WER reduction is experimentally obtained and the numerical simulationsThe inverse reproduce bias method it, as shown is another in Figures unique 20d technique and 20e. for reducing the WER. Figure 21a illustrates the write sequence of the conventional and inverse bias methods. In the inverse bias method, a bias voltage with a negative polarity is applied before and after the write pulse. If the system exhibits a linear VCMA effect, then the inverse bias enhances the KPMA of the free layer, and thereby reduces the thermal fluctuations in the initial state and during the relaxation process. It should be noted that inverse biases can also be used for the pre-read and read verify processes, which thereby offers a read-disturbance-free operation as well as WER reduction. Noguchi et al. first proposed the inverse bias method [37] and the effectiveness was later studied using numerical simulations [144]. In Ref. 144, a substantial reduction in WER was confirmed by introducing inverse biases, whose absolute intensity was the same as that of the write pulse, but with opposite sign (see Figure 21b).

MicromachinesMicromachines2019 ,201910, 327, 10, x 2222 of of 31 31

Micromachines 2019, 10, x 23 of 31 Figure 20. (a)–(c)Effects of pulse shaping on magnetization trajectory. The red and green lines represent Ω current,the magnetizationFigure whereas 20. ( anearly)–( trajectoryc) Effects all micr during of owavepulse and shaping interconnects after application on magnetizat have of the aion characteristic pulse, trajectory.tpulse, respectively.The impedance red and ( dgreenof), (50e) WERlines. This impedanceminimumrepresent mismatch as a the function magnetization gives of rise rise time to trajectory multiple (blue symbols) duri reflectionsng and after fall between time application (red the symbols). signal of the source pulse, (d) experimental andtpulse ,the respectively. MTJ, results; and/or the(e )deformation numerical(d), (e) WER simulation of minimum the waveform, results. as a function Reprinted and this of figurerise obscur time withes (blue permissionthe symbols)correlation from and [between 108fall], time Copyright the(red appliedsymbols). 2019 by voltage ( thed) waveformAmericanexperimental and Physical the inducedresults; Society. (e magnetization) numerical simulation dynamics. results. Reprinted figure with permission from [108], Copyright 2019 by the American Physical Society.

The inverse bias method is another unique technique for reducing the WER. Figure 21a illustrates the write sequence of the conventional and inverse bias methods. In the inverse bias method, a bias voltage with a negative polarity is applied before and after the write pulse. If the system exhibits a linear VCMA effect, then the inverse bias enhances the KPMA of the free layer, and thereby reduces the thermal fluctuations in the initial state and during the relaxation process. It should be noted that inverse biases can also be used for the pre-read and read verify processes, which thereby offers a read-disturbance-free operation as well as WER reduction. Noguchi et al. first proposed the inverse bias method [37] and the effectiveness was later studied using numerical simulations [144]. In Ref. 144, a substantial reduction in WER was confirmed by introducing inverse biases, whose absolute intensity was the same as that of the write pulse, but with opposite sign (see Figure 21 b). Since precise control of voltage-driven magnetization switching relies on the precise control of the voltage pulse shape, accurate calculation and shaping of the voltage pulse waveform [145,38] are also an important technique for studying the voltage-driven magnetization dynamics in detail. The procedure that is presented in Ref. 145 allows for one to accurately analyze and control the voltage FigureFigure 21. 21.( a(a)) ComparisonComparison of of write write pulse pulse sequence sequence in inconventional conventional and andinverse inverse bias methods. bias methods. (b) waveform applied to an MTJ. This is especially important in the development of voltage-torque (b)Numerically Numerically obtained obtained WER WER as asa function a function of Δ of using∆ using two two different different methods. methods. MRAM, because the MTJ resistance becomes much higher than 50 Ω to suppress the flow of charge An external bias magnetic field has been used to determine the axis for magnetization precession

in most experimental demonstrations of voltage-induced dynamic switching. However, the application of a magnetic field is not suitable for practical circuits. Therefore, we also need efforts to replace the external bias field by an effective field, such as through crystalline anisotropy and exchange bias fields. Matsumoto et al. proposed using a combination of a conical magnetization state and shape anisotropy to induce precessional switching under zero-bias magnetic field [146]. Conical magnetization states have been mainly studied in multilayer structures containing Co, such as Co/Pt and Co/Pd [147–149], however recently it can be realized, even in a practical CoFeB/MgO structure [150–152]. Therefore, the above proposed structure might be applicable if we can realize a sufficiently-high thermal stability while keeping the conical states.

6. Conclusions Electric-field control of spin has the potential to make substantial impact on the development of novel nonvolatile memory with ultra-low operating power, as well as the expected zero stand-by power. The utilization of the voltage-controlled magnetic anisotropy (VCMA) effect is a promising approach to realizing voltage-torque MRAMs. Bi-stable magnetization switching has been demonstrated while using precessional dynamics that are induced by the VCMA effect. The purely- electronic VCMA effect originates from electric-field induced modification of the electronic structure at the interface between an ultrathin ferromagnet and a dielectric layer, such as MgO. In a 3d transition ferromagnet, e.g. Fe and Co, the voltage-induced change in the orbital magnetic moment plays an important role in the origin of the VCMA effect through the carrier accumulation/depletion effect at the interface. On the other hand, in a 3d/5d composite system, e.g. L10-FePt film, an electric quadrupole mechanism also has significant influence on the VCMA effect. To increase of the VCMA coefficient, the utilization of proximity-induced magnetism in a 5d transition metal, which has large spin-orbit coupling, is promising. A large VCMA coefficient of −320 fJ/Vm has been achieved in an Ir-doped ultrathin Fe layer with a demonstration of high-speed responsiveness. As for the reliability of writing while using voltage-induced dynamic switching, low write error rates of the order of 10-6 have been realized by improving the thermal stability and the VCMA effect in practical perpendicularly-magnetized MTJs. Further enhancement in the VCMA coefficient is the key to

Micromachines 2019, 10, 327 23 of 31

Since precise control of voltage-driven magnetization switching relies on the precise control of the voltage pulse shape, accurate calculation and shaping of the voltage pulse waveform [38,145] are also an important technique for studying the voltage-driven magnetization dynamics in detail. The procedure that is presented in Ref. 145 allows for one to accurately analyze and control the voltage waveform applied to an MTJ. This is especially important in the development of voltage-torque MRAM, because the MTJ resistance becomes much higher than 50 Ω to suppress the flow of charge current, whereas nearly all microwave interconnects have a characteristic impedance of 50 Ω. This impedance mismatch gives rise to multiple reflections between the signal source and the MTJ, and/or the deformation of the waveform, and this obscures the correlation between the applied voltage waveform and the induced magnetization dynamics. An external bias magnetic field has been used to determine the axis for magnetization precession in most experimental demonstrations of voltage-induced dynamic switching. However, the application of a magnetic field is not suitable for practical circuits. Therefore, we also need efforts to replace the external bias field by an effective field, such as through crystalline anisotropy and exchange bias fields. Matsumoto et al. proposed using a combination of a conical magnetization state and shape anisotropy to induce precessional switching under zero-bias magnetic field [146]. Conical magnetization states have been mainly studied in multilayer structures containing Co, such as Co/Pt and Co/Pd [147–149], however recently it can be realized, even in a practical CoFeB/MgO structure [150–152]. Therefore, the above proposed structure might be applicable if we can realize a sufficiently-high thermal stability while keeping the conical states.

6. Conclusions Electric-field control of spin has the potential to make substantial impact on the development of novel nonvolatile memory with ultra-low operating power, as well as the expected zero stand-by power. The utilization of the voltage-controlled magnetic anisotropy (VCMA) effect is a promising approach to realizing voltage-torque MRAMs. Bi-stable magnetization switching has been demonstrated while using precessional dynamics that are induced by the VCMA effect. The purely-electronic VCMA effect originates from electric-field induced modification of the electronic structure at the interface between an ultrathin ferromagnet and a dielectric layer, such as MgO. In a 3d transition ferromagnet, e.g. Fe and Co, the voltage-induced change in the orbital magnetic moment plays an important role in the origin of the VCMA effect through the carrier accumulation/depletion effect at the interface. On the other hand, in a 3d/5d composite system, e.g. L10-FePt film, an electric quadrupole mechanism also has significant influence on the VCMA effect. To increase of the VCMA coefficient, the utilization of proximity-induced magnetism in a 5d transition metal, which has large spin-orbit coupling, is promising. A large VCMA coefficient of 320 fJ/Vm has been achieved in an Ir-doped ultrathin Fe layer with a demonstration − of high-speed responsiveness. As for the reliability of writing while using voltage-induced dynamic 6 switching, low write error rates of the order of 10− have been realized by improving the thermal stability and the VCMA effect in practical perpendicularly-magnetized MTJs. Further enhancement in the VCMA coefficient is the key to demonstrating the potential for scalability and realizing more reliable switching for voltage-torque MRAM. A novel nonvolatile memory maintaining low operating power as well as zero stand-by power can provide a broader option for the design of memory hierarchy in future data-driven society. We expect that the voltage-torque MRAM has the potential to be applied in IoT edge devices and wearable/implantable computing systems, in which, ultimately, low power consumption is strongly demanded. Furthermore, the voltage-control of spin may also lead to the improvement in other spintronic devices, such as a voltage-tuned magnetic sensor, spin-torque oscillator, and spin-based neuromorphic devices.

Author Contributions: T.N. wrote Section1 “Introduction”, Section2 “Overview of the VCMA e ffect and voltage-induced dynamic switching”, and Section4 “Materials research for large VCMA e ffect”. T.Y. wrote Section5 “Towards the reliable voltage-induced dynamic switching”. S.M. wrote Section3 “Physical origin of the Micromachines 2019, 10, 327 24 of 31

VCMA effect”. M.T. wrote part of the theoretical discussion in Section4. M.S., Y.S. and S.Y. supervised the work and edited the manuscript. Acknowledgments: This work was supported by the ImPACT Program of the Council for Science. We thank Y. Shiota, A. Kozioł-Rachwał, W. Skowro´nski,X. Xu, T. Ikeura, T. Ohkubo, T. Tsukahara, M. Suzuki, S. Tamaru, H. Kubota, A. Fukushima, K. Hono, K. Nakamura, T. Oda, R. Matsumoto, H. Imamura, Y. Miura, T. Taniguchi, T. Yorozu, Y. Kotani, T. Nakamura and M. Sahashi for fruitful discussions. The XAS and XMCD measurements were performed in SPring-8 with the approval of the Japan Synchrotron Radiation Research Institute (Proposal Nos. 2016B1017). Conflicts of Interest: The authors declare no conflict of interest.

References

1. Yuasa, S.; Nagahama, T.; Fukushima, A.; Suzuki, Y.; Ando, K. Giant room-temperature magnetoresistance in single-crystal Fe/MgO/Fe magnetic tunnel junctions. Nat. Mater. 2004, 3, 868–871. [CrossRef] 2. Parkin, S.S.; Kaiser, C.; Panchula, A.; Rice, P.M.; Hughes, B.; Samant, M.; Yang, S.H. Giant tunnelling magnetoresistance at room temperature with MgO (100) tunnel barriers. Nat. Mater. 2004, 3, 862–867. [CrossRef] 3. Slonczewski, J.C. Current-driven excitation of magnetic multilayers. J. Magn. Magn. Mater. 1996, 159, L1. [CrossRef] 4. Berger, L. Emission of spin waves by a magnetic multilayer traversed by a current. Phys. Rev. B 1996, 54, 9353–9358. [CrossRef] 5. Myers, E.B.; Ralph, D.C.; Katine, J.A.; Louie, R.N.; Buhrman, R.A. Current-induced switching of domains in magnetic multilayer devices. Science 1999, 285, 867–870. [CrossRef][PubMed] 6. Katine, J.A.; Albert, F.J.; Buhrman, R.A.; Myers, E.B.; Ralph, D.C. Current-driven magnetization reversal and spin-wave excitation in Co/Cu/Co pillars. Phys. Rev. Lett. 2000, 84, 3149–3152. [CrossRef] 7. Huai, Y.; Albert, F.; Nguyen, P.; Pakala, M.; Valet, T. Observation of spin-transfer switching in deep submicron-sized and low-resistance magnetic tunnel junctions. Appl. Phys. Lett. 2004, 84, 3118–3120. [CrossRef] 8. Kubota, H.; Fukushima, A.; Ootani, Y.; Yuasa, S.; Ando, K.; Maehara, H.; Tsunekawa, K.; Djayaprawira, D.D.; Watanabe, N.; Suzuki, Y. Evaluation of Spin-Transfer Switching in CoFeB/MgO/CoFeB Magnetic Tunnel Junctions. Jpn. J. Appl. Phys. 2005, 44, L1237. [CrossRef] 9. Chernyshov, A.; Overby, M.; Liu, X.; Furdyna, J.K.; Lyanda-Geller, Y.; Rokhinson, L.P. Evidence for reversible control of magnetization in a ferromagnetic material by means of spin–orbit magnetic field. Nat. Phys. 2009, 5, 656–659. [CrossRef] 10. Miron, I.M.; Gaudin, G.; Auffret, S.; Rodmacq, B.; Schuhl, A.; Pizzini, S.; Vogel, J.; Gambardella, P. Current-driven spin torque induced by the Rashba effect in a ferromagnetic metal layer. Nat. Mater. 2010, 9, 230–234. [CrossRef] 11. Miron, I.M.; Garello, K.; Gaudin, G.; Zermatten, P.J.; Costache, M.V.; Auffret, S.; Bandiera, S.; Rodmacq, B.; Schuhl, A.; Gambardella, P. Perpendicular switching of a single ferromagnetic layer induced by in-plane current injection. Nature 2011, 476, 189–193. [CrossRef][PubMed] 12. Liu, L.; Pai, C.-F.; Li, Y.; Tseng, H.W.; Ralph, D.C.; Buhrman, R.A. Spin-torque switching with the giant spin Hall effect of tantalum. Science 2012, 336, 555–558. [CrossRef][PubMed] 13. Ando, K.; Fujita, S.; Ito, J.; Yuasa, S.; Suzuki, Y.; Nakatani, Y.; Miyazaki, T.; Yoda, H. Spin-transfer torque magnetoresistive random-access memory technologies for normally off computing (invited). J. Appl. Phys. 2014, 115, 172607. [CrossRef] 14. Khvalkovskiy, A.V.; Apalkov, D.; Watts, S.; Chepulskii, R.; Beach, R.S.; Ong, A.; Tang, X.; Driskill-Smith, A.; Butler, W.H.; Visscher, B.P.; et al. Basic principles of STT-MRAM cell operation in memory arrays. J. Phys. D Appl. Phys. 2013, 46, 074001. [CrossRef] 15. Rizal, C.; Moa, B.; Niraula, B. Ferromagnetic Multilayers: Magnetoresistance, Magnetic Anisotropy, and Beyond. Magnetochemistry 2016, 2, 22. [CrossRef] 16. Novosad, V.; Otani, Y.; Ohsawa, A.; Kim, S.G.; Fukamichi, K.; Koike, J.; Maruyama, K.; Kitakami, O.; Shimada, Y. Novel magnetostrictive memory device. J. Appl. Phys. 2000, 87, 6400–6402. [CrossRef] 17. Hu, J.M.; Li, Z.; Chen, L.Q.; Nan, C.W. High-density magnetoresistive random access memory operating at ultralow voltage at room temperature. Nat. Commun. 2011, 2, 553. [CrossRef][PubMed] Micromachines 2019, 10, 327 25 of 31

18. Wu, T.; Bur, A.; Wong, K.; Zhao, P.; Lynch, C.S.; Amiri, P.K.; Wang, K.L.; Carman, G.P. Electrical control of reversible and permanent magnetization reorientation for magnetoelectric memory devices. Appl. Phys. Lett. 2011, 98, 262504. 19. Ohno, H.; Chiba, D.; Matsukura, F.; Omiya, T.; Abe, E.; Dietl, T.; Ohno, Y.; Ohtani, K. Electric-field control of ferromagnetism. Nature 2000, 408, 944–946. [CrossRef][PubMed] 20. Chiba, D.; Yamanouchi, M.; Matsukura, F.; Ohno, H. Electrical manipulation of magnetization reversal in a ferromagnetic semiconductor. Science 2003, 301, 943–945. [CrossRef][PubMed] 21. Yamada, Y.; Ueno, K.; Fukumura, T.; Yuan, H.T.; Shimotani, H.; Iwasa, Y.; Gu, L.; Tsukimoto, S.; Ikuhara, Y.; Kawasaki, M. Electrically induced ferromagnetism at room temperature in cobalt-doped titanium dioxide. Science 2011, 332, 1065–1067. [CrossRef][PubMed] 22. Chiba, D.; Fukami, S.; Shimamura, K.; Ishiwata, N.; Kobayashi, K.; Ono, T. Electrical control of the ferromagnetic phase transition in cobalt at room temperature. Nat. Mater. 2011, 10, 853–856. [CrossRef] [PubMed] 23. Borisov, P.; Hochstrat, A.; Chen, X.; Kleemann, W.; Binek, C. Magnetoelectric switching of exchange bias. Phys. Rev. Lett. 2005, 94, 117203. [CrossRef][PubMed] 24. He, X.; Wang, Y.; Wu, N.; Caruso, A.N.; Vescovo, E.; Belashchenko, K.D.; Dowben, P.A.; Binek, C. Robust isothermal electric control of exchange bias at room temperature. Nat. Mater. 2010, 9, 579–585. [CrossRef] [PubMed] 25. Ashida, T.; Oida, M.; Shimomura, N.; Nozaki, T.; Shibata, T.; Sahashi, M. Observation of magnetoelectric

effect in Cr2O3/Pt/Co thin film system. Appl. Phys. Lett. 2014, 104, 152409. [CrossRef] 26. Toyoki, K.; Shiratsuchi, Y.; Kobane, A.; Mitsumata, C.; Kotani, Y.; Nakamura, T.; Nakatani, R. Magnetoelectric

switching of perpendicular exchange bias in Pt/Co/α-Cr2O3/Pt stacked films. Appl. Phys. Lett. 2015, 106, 162404. [CrossRef]

27. Duan, C.G.; Jaswal, S.S.; Tsymbal, E.Y. Predicted magnetoelectric effect in Fe/BaTiO3 multilayers: Ferroelectric control of magnetism. Phys. Rev. Lett. 2006, 97, 047201. [CrossRef] 28. Radaelli, G.; Petti, D.; Plekhanov, E.; Fina, I.; Torelli, P.; Salles, B.R.; Cantoni, M.; Rinaldi, C.; Gutierrez, D.;

Panaccione, G.; et al. Electric control of magnetism at the Fe/BaTiO3 interface. Nat. Commun. 2014, 5, 3404. [CrossRef] 29. Gerhard, L.; Yamada, T.K.; Balashov, T.; Takacs, A.F.; Wesselink, R.J.; Dane, M.; Fechner, M.; Ostanin, S.; Ernst, A.; Mertig, I.; et al. Magnetoelectric coupling at metal surfaces. Nat. Nanotechnol. 2010, 5, 792–797. [CrossRef] 30. Heron, J.T.; Bosse, J.L.; He, Q.; Gao, Y.; Trassin, M.; Ye, L.; Clarkson, J.D.; Wang, C.; Liu, J.; Salahuddin, S.; et al. Deterministic switching of ferromagnetism at room temperature using an electric field. Nature 2014, 516, 370–373. [CrossRef][PubMed] 31. Tokunaga, Y.; Taguchi, Y.; Arima, T.-H.; Tokura, Y. Electric-field-induced generation and reversal of ferromagnetic moment in ferrites. Nat. Phys. 2012, 8, 838–844. [CrossRef] 32. Weisheit, M.; Fahler, S.; Marty, A.; Souche, Y.; Poinsignon, C.; Givord, D. Electric field-induced modification of magnetism in thin-film ferromagnets. Science 2007, 315, 349–351. [CrossRef][PubMed] 33. Maruyama, T.; Shiota, Y.; Nozaki, T.; Ohta, K.; Toda, N.; Mizuguchi, M.; Tulapurkar, A.A.; Shinjo, T.; Shiraishi, M.; Mizukami, S.; et al. Large voltage-induced magnetic anisotropy change in a few atomic layers of iron. Nat. Nanotechnol. 2009, 4, 158–161. [CrossRef] 34. Amiri, P.K.; Alzate, J.G.; Cai, X.Q.; Ebrahimi, F.; Hu, Q.; Wong, K.; Grezes, C.; Lee, H.; Yu, G.; Li, X.; et al. Electric-Field-Controlled Magnetoelectric RAM: Progress, Challenges, and Scaling. IEEE Trans. Magn. 2015, 51, 3401507. 35. Nozaki, T.; Kozioł-Rachwał, A.; Tsujikawa, M.; Shiota, Y.; Xu, X.; Ohkubo, T.; Tsukahara, T.; Miwa, S.; Suzuki, M.; Tamaru, S.; et al. Highly efficient voltage control of spin and enhanced interfacial perpendicular magnetic anisotropy in iridium-doped Fe/MgO magnetic tunnel junctions. NPG Asia Mater. 2017, 9, e451. [CrossRef] 36. Wang, K.L.; Lee, H.; Amiri, P.K. Magnetoelectric Random Access Memory-Based Circuit Design by Using Voltage-Controlled Magnetic Anisotropy in Magnetic Tunnel Junctions. IEEE Trans. Nanotechnol. 2015, 14, 992–997. [CrossRef] Micromachines 2019, 10, 327 26 of 31

37. Noguchi, H.; Ikegami, K.; Abe, K.; Fujita, S.; Shiota, Y.; Nozaki, T.; Yuasa, S.; Suzuki, Y. Novel Voltage Controlled MRAM(VCM) with Fast Read/Write Circuits for Ultra Large Last Level Cache. In Proceedings of the 2016 IEEE International Electron Devices Meeting (IEDM), San Francisco, CA, USA, 3–7 Decembder 2016. 38. Lee, H.; Lee, A.; Wang, S.; Ebrahimi, F.; Gupta, P.; Amiri, P.K.; Wang, K.L. Analysis and Compact Modeling of Magnetic Tunnel Junctions Utilizing Voltage-Controlled Magnetic Anisotropy. IEEE Trans. Magn. 2018, 54, 4400209. [CrossRef] 39. Long, M.; Zeng, L.; Gao, T.; Zhang, D.; Qin, X.; Zhang, Y.; Zhao, W. Self-Adaptive Write Circuit for Magnetic Tunneling Junction Memory With Voltage-Controlled Magnetic Anisotropy Effect. IEEE Trans. Nanotechnol. 2018, 17, 492–499. [CrossRef] 40. Chiba, D.; Sawicki, M.; Nishitani, Y.; Nakatani, Y.; Matsukura, F.; Ohno, H. Magnetization vector manipulation by electric fields. Nature 2008, 455, 515–518. [CrossRef] 41. Duan, C.G.; Velev, J.P.; Sabirianov, R.F.; Zhu, Z.; Chu, J.; Jaswal, S.S.; Tsymbal, E.Y. Surface magnetoelectric effect in ferromagnetic metal films. Phys. Rev. Lett. 2008, 101, 137201. [CrossRef] 42. Nakamura, K.; Shimabukuro, R.; Fujiwara, Y.; Akiyama, T.; Ito, T.; Freeman, A.J. Giant modification of the magnetocrystalline anisotropy in transition-metal monolayers by an external electric field. Phys. Rev. Lett. 2009, 102, 187201. [CrossRef][PubMed] 43. Tsujikawa, M.; Oda, T. Finite electric field effects in the large perpendicular magnetic anisotropy surface Pt/Fe/Pt(001): A first-principles study. Phys. Rev. Lett. 2009, 102, 247203. [CrossRef][PubMed] 44. Xu, L.; Zhang, S. Electric field control of interface magnetic anisotropy. J. Appl. Phys. 2012, 111, 07C501. [CrossRef] 45. Barnes, S.E.; Ieda, J.; Maekawa, S. Rashba spin-orbit anisotropy and the electric field control of magnetism. Sci. Rep. 2014, 4, 4105. [CrossRef][PubMed] 46. Nakamura, K.; Akiyama, T.; Ito, T.; Weinert, M.; Freeman, A.J. Role of an interfacial FeO layer in the electric-field-driven switching of magnetocrystalline anisotropy at the Fe/MgO interface. Phys. Rev. B 2010, 81, 220409(R). [CrossRef] 47. Shiota, Y.; Maruyama, T.; Nozaki, T.; Shinjo, T.; Shiraishi, A.M.; Suzuki, Y. Voltage-Assisted Magnetization

Switching in Ultrathin Fe80Co20Alloy Layers. Appl. Phys. Exp. 2009, 2, 063001. [CrossRef] 48. Nozaki, T.; Shiota, Y.; Shiraishi, M.; Shinjo, T.; Suzuki, Y. Voltage-induced perpendicular magnetic anisotropy change in magnetic tunnel junctions. Appl. Phy. Lett. 2010, 96, 022506. [CrossRef] 49. Shiota, Y.; Murakami, S.; Bonell, F.; Nozaki, T.; Shinjo, T.; Suzuki, Y. Quantitative Evaluation of Voltage-Induced Magnetic Anisotropy Change by Magnetoresistance Measurement. Appl. Phys. Exp. 2011, 4, 043005. [CrossRef] 50. Nozaki, T.; Shiota, Y.; Miwa, S.; Murakami, S.; Bonell, F.; Ishibashi, S.; Kubota, H.; Yakushiji, K.; Saruya, T.; Fukushima, A.; et al. Electric-field-induced ferromagnetic resonance excitation in an ultrathin ferromagnetic metal layer. Nat. Phys. 2012, 8, 492–496. [CrossRef] 51. Zhu, J.; Katine, J.A.; Rowlands, G.E.; Chen, Y.J.; Duan, Z.; Alzate, J.G.; Upadhyaya, P.; Langer, J.; Amiri, P.K.; Wang, K.L.; et al. Voltage-induced ferromagnetic resonance in magnetic tunnel junctions. Phys Rev Lett 2012, 108, 197203. [CrossRef] 52. Shiota, Y.; Miwa, S.; Tamaru, S.; Nozaki, T.; Kubota, H.; Fukushima, A.; Suzuki, Y.; Yuasa, S. High-output microwave detector using voltage-induced ferromagnetic resonance. Appl. Phys. Lett. 2014, 105, 192408. [CrossRef] 53. Kanai, S.; Gajek, M.; Worledge, D.C.; Matsukura, F.; Ohno, H. Electric field-induced ferromagnetic resonance in a CoFeB/MgO magnetic tunnel junction under dc bias voltages. Appl. Phys. Lett. 2014, 105, 242409. [CrossRef] 54. Rana, B.; Fukuma, Y.; Miura, K.; Takahashi, H.; Otani, Y. Effect of excitation power on voltage induced local magnetization dynamics in an ultrathin CoFeB film. Sci. Rep. 2017, 7, 2318. [CrossRef][PubMed] 55. Shiota, Y.; Nozaki, T.; Bonell, F.; Murakami, S.; Shinjo, T.; Suzuki, Y. Induction of coherent magnetization switching in a few atomic layers of FeCo using voltage pulses. Nat. Mater. 2012, 11, 39–43. [CrossRef] [PubMed] 56. Verba, R.; Carpentieri, M.; Finocchio, G.; Tiberkevich, V.; Slavin, A. Excitation of Spin Waves in an In-Plane-Magnetized Ferromagnetic Nanowire Using Voltage-Controlled Magnetic Anisotropy. Phys. Rev. Appl. 2017, 7, 064023. [CrossRef] Micromachines 2019, 10, 327 27 of 31

57. Chen, Y.J.; Lee, H.K.; Verba, R.; Katine, J.A.; Barsukov, I.; Tiberkevich, V.; Xiao, J.Q.; Slavin, A.N.; Krivorotov, I.N. Parametric Resonance of Magnetization Excited by Electric Field. Nano Lett. 2017, 17, 572–577. [CrossRef][PubMed] 58. Rana, B.; Fukuma, Y.; Miura, K.; Takahashi, H.; Otani, Y. Excitation of coherent propagating spin waves in ultrathin CoFeB film by voltage-controlled magnetic anisotropy. Appl. Phys. Lett. 2017, 111, 052404. [CrossRef] 59. Bonell, F.; Murakami, S.; Shiota, Y.; Nozaki, T.; Shinjo, T.; Suzuki, Y. Large change in perpendicular magnetic anisotropy induced by an electric field in FePd ultrathin films. Appl. Phys. Lett. 2011, 98, 232510. [CrossRef] 60. Seki, T.; Kohda, M.; Nitta, J.; Takanashi, K. Coercivity change in an FePt thin layer in a Hall device by voltage application. Appl. Phys. Lett. 2011, 98, 212505. [CrossRef] 61. Kikushima, S.; Seki, T.; Uchida, K.; Saitoh, E.; Takanashi, K. Electric field effect on magnetic anisotropy for Fe-Pt-Pd alloys. AIP Adv. 2017, 7, 085210. [CrossRef] 62. Kozioł-Rachwał, A.; Nozaki, T.; Freindl, K.; Korecki, J.; Yuasa, S.; Suzuki, Y. Enhancement of perpendicular magnetic anisotropy and its electric field-induced change through interface engineering in Cr/Fe/MgO. Sci. Rep. 2017, 7, 5993. [CrossRef][PubMed] 63. Xiang, Q.; Wen, Z.; Sukegawa, H.; Kasai, S.; Seki, T.; Kubota, T.; Takanashi, K.; Mitani, S. Nonlinear electric field effect on perpendicular magnetic anisotropy in Fe/MgO interfaces. J. Phys. D Appl. Phys. 2017, 50, 40LT04. [CrossRef] 64. Wen, Z.; Sukegawa, H.; Seki, T.; Kubota, T.; Takanashi, K.; Mitani, S. Voltage control of magnetic anisotropy

in epitaxial Ru/Co2FeAl/MgO heterostructures. Sci. Rep. 2017, 7, 45026. [CrossRef][PubMed] 65. Miwa, S.; Fujimoto, J.; Risius, P.; Nawaoka, K.; Goto, M.; Suzuki, Y. Strong Bias Effect on Voltage-Driven Torque at Epitaxial Fe-MgO Interface. Phys. Rev. X 2017, 7, 031018. [CrossRef] 66. Shukla, A.K.; Goto, M.; Xu, X.; Nawaoka, K.; Suwardy, J.; Ohkubo, T.; Hono, K.; Miwa, S.; Suzuki, Y. Voltage-Controlled Magnetic Anisotropy in Fe1-xCox/Pd/MgO system. Sci. Rep. 2018, 8, 10362. [CrossRef] [PubMed] 67. Suzuki, K.Z.; Kimura, S.; Kubota, H.; Mizukami, S. Magnetic Tunnel Junctions with a Nearly Zero Moment Manganese Nanolayer with Perpendicular Magnetic Anisotropy. ACS Appl. Mater. Interfaces 2018, 10, 43305. [CrossRef][PubMed] 68. Endo, M.; Kanai, S.; Ikeda, S.; Matsukura, F.; Ohno, H. Electric-field effects on thickness dependent magnetic

anisotropy of sputtered MgO/Co40Fe40B20/Ta structures. Appl. Phys. Lett. 2010, 96, 212503. [CrossRef] 69. Kita, K.; Abraham, D.W.; Gajek, M.J.; Worledge, D.C. Electric-field-control of magnetic anisotropy of Co0.6Fe0.2B0.2/oxide stacks using reduced voltage. J. Appl. Phys. 2012, 112, 033919. [CrossRef] 70. Skowron´nski,W.; Wis´sniowski,P.; Stobiecki, T.; Cardoso, S.; Freitas, P.P.; van Dijken, S. Magnetic field sensor with voltage-tunable sensing properties. Appl. Phys. Lett. 2012, 101, 192401. [CrossRef] 71. Nozaki, T.; Yakushiji, K.; Tamaru, S.; Sekine, M.; Matsumoto, R.; Konoto, M.; Kubota, H.; Fukushima, A.; Yuasa, S. Voltage-Induced Magnetic Anisotropy Changes in an Ultrathin FeB Layer Sandwiched between Two MgO Layers. Appl. Phys. Exp. 2013, 6, 073005. [CrossRef] 72. Shiota, Y.; Bonell, F.; Miwa, S.; Mizuochi, N.; Shinjo, T.; Suzuki, Y. Opposite signs of voltage-induced perpendicular magnetic anisotropy change in CoFeB MgO junctions with different underlayers. Appl. Phys. | Lett. 2013, 103, 082410. [CrossRef] 73. Alzate, J.G.; Amiri, P.K.; Yu, G.; Upadhyaya, P.; Katine, J.A.; Langer, J.; Ocker, B.; Krivorotov, I.N.; Wang, K.L. Temperature dependence of the voltage-controlled perpendicular anisotropy in nanoscale MgO|CoFeB|Ta magnetic tunnel junctions. Appl. Phys. Lett. 2014, 104, 112410. [CrossRef] 74. Meng, H.; Naik, V.B.; Liu, R.; Han, G. Electric field control of spin re-orientation in perpendicular magnetic tunnel junctions—CoFeB and MgO thickness dependence. Appl. Phys. Lett. 2014, 105, 042410. [CrossRef] 75. Okada, A.; Kanai, S.; Yamanouchi, M.; Ikeda, S.; Matsukura, F.; Ohno, H. Electric-field effects on magnetic anisotropy and damping constant in Ta/CoFeB/MgO investigated by ferromagnetic resonance. Appl. Phys. Lett. 2014, 105, 052415. [CrossRef] 76. Li, X.; Yu, G.; Wu, H.; Ong, P.V.; Wong, K.; Hu, Q.; Ebrahimi, F.; Upadhyaya, P.; Akyol, M.; Kioussis, N.; et al. Thermally stable voltage-controlled perpendicular magnetic anisotropy in Mo|CoFeB|MgO structures. Appl. Phys. Lett. 2015, 107, 142403. [CrossRef] Micromachines 2019, 10, 327 28 of 31

77. Skowro´nski,W.; Nozaki, T.; Lam, D.D.; Shiota, Y.; Yakushiji, K.; Kubota, H.; Fukushima, A.; Yuasa, S.; Suzuki, Y. Underlayer material influence on electric-field controlled perpendicular magnetic anisotropy in CoFeB/MgO magnetic tunnel junctions. Phys. Rev. B 2015, 91, 184410. [CrossRef] 78. Skowro´nski,W.; Nozaki, T.; Shiota, Y.; Tamaru, S.; Yakushiji, K.; Kubota, H.; Fukushima, A.; Yuasa, S.; Suzuki, Y. Perpendicular magnetic anisotropy of Ir/CoFeB/MgO trilayer system tuned by electric fields. Appl. Phys. Exp. 2015, 8, 053003. [CrossRef] 79. Piotrowski, S.K.; Bapna, M.; Oberdick, S.D.; Majetich, S.A.; Li, M.; Chien, C.L.; Ahmed, R.; Victora, R.H. Size and voltage dependence of effective anisotropy in sub-100-nm perpendicular magnetic tunnel junctions. Phys. Rev. B 2016, 94, 014404. [CrossRef] 80. YLau, -C.; Sheng, P.; Mitani, S.; Chiba, D.; Hayashi, M. Electric field modulation of the non-linear areal magnetic anisotropy energy. Appl. Phys. Lett. 2017, 110, 022405. 81. Li, X.; Fitzell, K.; Wu, D.; Karaba, C.T.; Buditama, A.; Yu, G.; Wong, K.L.; Altieri, N.; Grezes, C.; Kioussis, N.; et al. Enhancement of voltage-controlled magnetic anisotropy through precise control of Mg insertion thickness at CoFeB|MgO interface. Appl. Phys. Lett. 2017, 110, 052401. [CrossRef] 82. Sonntag, A.; Hermenau, J.; Schlenhoff, A.; Friedlein, J.; Krause, S.; Wiesendanger, R. Electric-field-induced magnetic anisotropy in a nanomagnet investigated on the atomic scale. Phys. Rev. Lett. 2014, 112, 017204. [CrossRef][PubMed] 83. Zhou, T.; Leong, S.H.; Yuan, Z.M.; Hu, S.B.; Ong, C.L.; Liu, B. Manipulation of magnetism by electrical field in a real recording system. Appl. Phys. Lett. 2010, 96, 012506. [CrossRef] 84. Bauer, U.; Przybylski, M.; Beach, G.S.D. Voltage control of magnetic anisotropy in Fe films with quantum well states. Phys. Rev. B 2014, 89, 174402. [CrossRef] 85. Chiba, D.; Kawaguchi, M.; Fukami, S.; Ishiwata, N.; Shimamura, K.; Kobayashi, K.; Ono, T. Electric-field control of magnetic domain-wall velocity in ultrathin cobalt with perpendicular magnetization. Nat. Commun. 2012, 3, 888. [CrossRef][PubMed] 86. Schellekens, A.J.; van den Brink, A.; Franken, J.H.; Swagten, H.J.; Koopmans, B. Electric-field control of domain wall motion in perpendicularly magnetized materials. Nat. Commun. 2012, 3, 847. [CrossRef] [PubMed] 87. Bauer, U.; Emori, S.; Beach, G.S.D. Voltage-controlled domain wall traps in ferromagnetic nanowires. Nat. Nanotechnol. 2013, 8, 411–416. [CrossRef][PubMed] 88. Schott, M.; Bernand-Mantel, A.; Ranno, L.; Pizzini, S.; Vogel, J.; Bea, H.; Baraduc, C.; Auffret, S.; Gaudin, G.; Givord, D. The Skyrmion Switch: Turning Magnetic Skyrmion Bubbles on and off with an Electric Field. Nano Lett. 2017, 17, 3006–3012. [CrossRef][PubMed] 89. Srivastava, T.; Schott, M.; Juge, R.; Krizakova, V.; Belmeguenai, M.; Roussigne, Y.; Bernand-Mantel, A.; Ranno, L.; Pizzini, S.; Cherif, S.M.; et al. Large-Voltage Tuning of Dzyaloshinskii-Moriya Interactions: A Route toward Dynamic Control of Skyrmion Chirality. Nano Lett. 2018, 18, 4871–4877. [CrossRef] [PubMed] 90. Nozaki, T.; Jibiki, Y.; Goto, M.; Tamura, E.; Nozaki, T.; Kubota, H.; Fukushima, A.; Yuasa, S.; Suzuki, Y. Brownian motion of skyrmion bubbles and its control by voltage applications. Appl. Phys. Lett. 2019, 114, 012402. [CrossRef] 91. Nawaoka, K.; Miwa, S.; Shiota, Y.; Mizuochi, N.; Suzuki, Y. Voltage induction of interfacial Dzyaloshinskii–Moriya interaction in Au/Fe/MgO artificial multilayer. Appl. Phys. Express 2015, 8, 063004. [CrossRef] 92. Newhouse-Illige, T.; Liu, Y.; Xu, M.; Hickey, D.R.; Kundu, A.; Almasi, H.; Bi, C.; Wang, X.; Freeland, J.W.; Keavney, D.J.; et al. Voltage-controlled interlayer coupling in perpendicularly magnetized magnetic tunnel junctions. Nat. Commun. 2017, 8, 15232. [CrossRef] 93. Hibino, Y.; Koyama, T.; Obinata, A.; Miwa, K.; Ono, S.; Chiba, D. Electric field modulation of magnetic anisotropy in perpendicularly magnetized Pt/Co structure with a Pd top layer. Appl. Phys. Express 2015, 8, 113002. [CrossRef] 94. Obinata, A.; Hibino, Y.; Hayakawa, D.; Koyama, T.; Miwa, K.; Ono, S.; Chiba, D. Electric-field control of magnetic moment in Pd. Sci. Rep. 2015, 5, 14303. [CrossRef] 95. Hibino, T.K.Y.; Obinata, A.; Harai, T.; Ota, S.; Miwa, K.; Ono, S.; Matsukura, F.; Ohno, H.; Chiba, D. Peculiar temperature dependence of electric-field effect on magnetic anisotropy in Co/Pd/MgO system. Appl. Phys. Lett. 2016, 109, 082403. [CrossRef] Micromachines 2019, 10, 327 29 of 31

96. Amiri, P.K.; Upadhyaya, P.; Alzate, J.G.; Wang, K.L. Electric-field-induced thermally assisted switching of monodomain magnetic bits. J. Appl. Phys. 2013, 113, 013912. [CrossRef] 97. Han, G.; Huang, J.; Chen, B.; Lim, S.T.; Tran, M. Electric Field Assisted Switching in Magnetic Random Access Memory. IEEE Trans. Magn. 2015, 51, 3401207. [CrossRef] 98. Nozaki, T.; Arai, H.; Yakushiji, K.; Tamaru, S.; Kubota, H.; Imamura, H.; Fukushima, A.; Yuasa, S. Magnetization switching assisted by high-frequency-voltage-induced ferromagnetic resonance. Appl. Phys. Express 2014, 7, 073002. [CrossRef] 99. Wang, W.G.; Li, M.; Hageman, S.; Chien, C.L. Electric-field-assisted switching in magnetic tunnel junctions. Nat. Mater. 2012, 11, 64–68. [CrossRef][PubMed] 100. Kanai, S.; Nakatani, Y.; Yamanouchi, M.; Ikeda, S.; Sato, H.; Matsukura, F.; Ohno, H. Magnetization switching in a CoFeB/MgO magnetic tunnel junction by combining spin-transfer torque and electric field-effect. Appl. Phys. Lett. 2014, 104, 212406. [CrossRef] 101. Yoda, N.S.H.; Ohsawa, Y.; Shirotori, S.; Kato, Y.; Inokuchi, T.; Kamiguchi, Y.; Altansargai, B.; Saito, Y.; Koi, K.; Sugiyama, H.; et al. Voltage-Control Spintronics Memory (VoCSM) Having Potentials of Ultra-Low Energy-Consumption and High-Density. In Proceedings of the 2016 IEEE International Electron Devices Meeting (IEDM), San Francisco, CA, USA, 3–7 December 2016. 102. Shiota, Y.; Miwa, S.; Nozaki, T.; Bonell, F.; Mizuochi, N.; Shinjo, T.; Kubota, H.; Yuasa, S.; Suzuki, Y. Pulse voltage-induced dynamic magnetization switching in magnetic tunneling junctions with high resistance-area product. Appl. Phys. Lett. 2012, 101, 102406. [CrossRef] 103. Kanai, S.; Yamanouchi, M.; Ikeda, S.; Nakatani, Y.; Matsukura, F.; Ohno, H. Electric field-induced magnetization reversal in a perpendicular-anisotropy CoFeB-MgO magnetic tunnel junction. Appl. Phys. Lett. 2012, 101, 122403. [CrossRef] 104. Grezes, C.; Rozas, A.R.; Ebrahimi, F.; Alzate, J.G.; Cai, X.; Katine, J.A.; Langer, J.; Ocker, B.; Amiri, P.K.; Wang, K.L. In-plane magnetic field effect on switching voltage and thermal stability in electric-field-controlled perpendicular magnetic tunnel junctions. AIP Adv. 2016, 6, 075014. [CrossRef] 105. Shiota, Y.; Nozaki, T.; Tamaru, S.; Yakushiji, K.; Kubota, H.; Fukushima, A.; Yuasa, S.; Suzuki, Y. Evaluation of write error rate for voltage-driven dynamic magnetization switching in magnetic tunnel junctions with perpendicular magnetization. Appl. Phys. Exp. 2016, 9, 013001. [CrossRef] 106. Shiota, Y.; Nozaki, T.; Tamaru, S.; Yakushiji, K.; Kubota, H.; Fukushima, A.; Yuasa, S.; Suzuki, Y. Reduction in write error rate of voltage-driven dynamic magnetization switching by improving thermal stability factor. Appl. Phys. Lett. 2017, 111, 022408. [CrossRef] 107. Yamamoto, T.; Nozaki, T.; Shiota, Y.; Imamura, H.; Tamaru, S.; Yakushiji, K.; Kubota, H.; Fukushima, A.; Suzuki, Y.; Yuasa, S. Thermally Induced Precession-Orbit Transition of Magnetization in Voltage-Driven Magnetization Switching. Phys. Rev. Appl. 2018, 10, 024004. [CrossRef] 108. Yamamoto, T.; Nozaki, T.; Imamura, H.; Shiota, Y.; Ikeura, T.; Tamaru, S.; Yakushiji, K.; Kubota, H.; Fukushima, A.; Suzuki, Y.; et al. Write-Error Reduction of Voltage-Torque-Driven Magnetization Switching by a Controlled Voltage Pulse. Phys. Rev. Appl. 2019, 11, 014013. [CrossRef] 109. Yamamoto, T.; Nozaki, T.; Imamura, H.; Shiota, Y.; Tamaru, S.; Yakushiji, K.; Kubota, H.; Fukushima, A.; Suzuki, Y.; Yuasa, S. Improvement of write error rate in voltage-driven magnetization switching. J. Phys. D Appl. Phys. 2019, 52, 164001. [CrossRef] 110. Grezes, C.; Lee, H.; Lee, A.; Wang, S.; Ebrahimi, F.; Li, X.; Wong, K.; Katine, J.A.; Ocker, B.; Langer, J.; et al. Write Error Rate and Read Disturbance in Electric-Field-Controlled MRAM. IEEE Magn. Lett. 2016, 8, 3102705. 111. Miwa, S.; Suzuki, M.; Tsujikawa, M.; Nozaki, T.; Nakamura, T.; Shirai, M.; Yuasa, S.; Suzuki, Y. Perpendicular magnetic anisotropy and its electric-field-induced change at metal-dielectric interfaces. J. Phys. D Appl. Phys. 2019, 52, 063001. [CrossRef] 112. Bruno, P. Tight-binding approach to the orbital magnetic moment and magnetocrystalline anisotropy of transition-metal monolayers. Phys. Rev. B 1989, 39, 865–868. [CrossRef] 113. Kawabe, T.; Yoshikawa, K.; Tsujikawa, M.; Tsukahara, T.; Nawaoka, K.; Kotani, Y.; Toyoki, K.; Goto, M.; Suzuki, M.; Nakamura, T.; et al. Electric-field-induced changes of magnetic moments and magnetocrystalline anisotropy in ultrathin cobalt films. Phys. Rev. B 2017, 96, 220412(R). [CrossRef] 114. Suzuki, Y.; Miwa, S. Magnetic anisotropy of ferromagnetic metals in low-symmetry systems. Phys. Lett. A 2019, 383, 1203–1206. [CrossRef] Micromachines 2019, 10, 327 30 of 31

115. Van der Laan, G. Microscopic origin of magnetocrystalline anisotropy in transition metal thin films. J. Phys. Condens. Mater. 1997, 10, 3239–3253. [CrossRef] 116. Miwa, S.; Suzuki, M.; Tsujikawa, M.; Matsuda, K.; Nozaki, T.; Tanaka, K.; Tsukahara, T.; Nawaoka, K.; Goto, M.; Kotani, Y.; et al. Voltage controlled interfacial magnetism through platinum orbits. Nat. Commun. 2017, 8, 15848. [CrossRef][PubMed] 117. Thole, B.T.; Carra, P.; Sette, F.; van der Laan, G. X-ray circular dichroism as a probe of orbital magnetization. Phys. Rev. Lett. 1992, 68, 1943–1946. [CrossRef][PubMed] 118. Carra, P.; Thole, B.T.; Altarelli, M.; Wang, X. X-ray circular dichroism and local magnetic fields. Phys. Rev. Lett. 1993, 70, 694–697. [CrossRef][PubMed] 119. Bonell, F.; Takahashi, Y.T.; Lam, D.D.; Yoshida, S.; Shiota, Y.; Miwa, S.; Nakamura, T.; Suzuki, Y. Reversible change in the oxidation state and magnetic circular dichroism of Fe driven by an electric field at the FeCo/MgO interface. Appl. Phys. Lett. 2013, 102, 152401. [CrossRef] 120. Miwa, S.; Matsuda, K.; Tanaka, K.; Kotani, Y.; Goto, M.; Nakamura, T.; Suzuki, Y. Voltage-controlled magnetic anisotropy in Fe|MgO tunnel junctions studied by x-ray absorption spectroscopy. Appl. Phys. Lett. 2015, 107, 162402. [CrossRef] 121. Suzuki, M.; Tsukahara, T.; Miyakaze, R.; Furuta, T.; Shimose, K.; Goto, M.; Nozaki, T.; Yuasa, S.; Suzuki, Y.; Miwa, S. Extended X-ray absorption fine structure analysis of voltage-induced effects in the interfacial atomic structure of Fe/Pt/MgO. Appl. Phys. Express 2017, 10, 063006. [CrossRef] 122. Bi, C.; Liu, Y.; Newhouse-Illige, T.; Xu, M.; Rosales, M.; Freeland, J.W.; Mryasov, O.; Zhang, S.; Velthuis, S.G.T.; Wang, W.G. Reversible control of Co magnetism by voltage-induced oxidation. Phys. Rev. Lett. 2014, 113, 267202. [CrossRef][PubMed] 123. Tsukahara, T.; Kawabe, T.; Shimose, K.; Furuta, T.; Miyakaze, R.; Nawaoka, K.; Goto, M.; Nozaki, T.; Yuasa, S.; Kotani, Y.; et al. Characterization of the magnetic moments of ultrathin Fe film in an external electric field via high-precision X-ray magnetic circular dichroism spectroscopy. Jpn. J. Appl. Phys. 2017, 56, 060304. [CrossRef] 124. Yamada, K.T.; Suzuki, M.; Pradipto, A.M.; Koyama, T.; Kim, S.; Kim, K.J.; Ono, S.; Taniguchi, T.; Mizuno, H.; Ando, F.; et al. Microscopic Investigation into the Electric Field Effect on Proximity-Induced Magnetism in Pt. Phys. Rev. Lett. 2018, 120, 157203. [CrossRef][PubMed] 125. Bauer, U.; Yao, L.; Tan, A.J.; Agrawal, P.; Emori, S.; Tuller, H.L.; van Dijken, S.; Beach, G.S. Magneto-ionic control of interfacial magnetism. Nat. Mater. 2015, 14, 174–181. [CrossRef] 126. Gilbert, D.A.; Grutter, A.J.; Arenholz, E.; Liu, K.; Kirby, B.J.; Borchers, J.A.; Maranville, B.B. Structural and magnetic depth profiles of magneto-ionic heterostructures beyond the interface limit. Nat. Commun. 2016, 7, 12264. [CrossRef][PubMed] 127. Sakamaki, M.; Amemiya, K. Observation of an electric field-induced interface redox reaction and magnetic modification in GdOx/Co thin film by means of depth-resolved X-ray absorption spectroscopy. Phys. Chem. Chem. Phys. 2018, 20, 20004. [CrossRef][PubMed] 128. Bauer, U.; Przybylski, M.; Kirschner, J.; Beach, G.S. Magnetoelectric charge trap memory. Nano Lett. 2012, 12, 1437–1442. [CrossRef][PubMed] 129. Rajanikanth, A.; Hauet, T.; Montaigne, F.; Mangin, S.; Andrieu, S. Magnetic anisotropy modified by electric field in V/Fe/MgO(001)/Fe epitaxial magnetic tunnel junction. Appl. Phys. Lett. 2013, 103, 062402. [CrossRef] 130. Ong, P.V.; Kioussis, N.; Odkhuu, D.; Amiri, P.K.; Wang, K.L.; Carman, G.P. Giant voltage modulation of magnetic anisotropy in strained heavy metal/magnet/insulator heterostructures. Phys. Rev. B 2015, 92, 020407(R). [CrossRef] 131. Odkhuu, D. Giant strain control of magnetoelectric effect in Ta|Fe|MgO. Sci. Rep. 2016, 6, 32742. [CrossRef] 132. Kato, Y.; Yoda, H.; Saito, Y.; Oikawa, S.; Fujii, K.; Yoshiki, M.; Koi, K.; Sugiyama, H.; Ishikawa, M.; Inokuchi, T.; et al. Giant voltage-controlled magnetic anisotropy effect in a crystallographically strained CoFe system. Appl. Phys. Express 2018, 11, 053007. [CrossRef] 133. Nozaki, T.; Kozioł-Rachwał, A.; Skowro´nski,W.; Zayets, V.; Shiota, Y.; Tamaru, S.; Kubota, H.; Fukushima, A.; Yuasa, S.; Suzuki, Y. Large Voltage-Induced Changes in the Perpendicular Magnetic Anisotropy of an MgO-Based Tunnel Junction with an Ultrathin Fe Layer. Phys. Rev. Appl. 2016, 5, 044006. [CrossRef] 134. Koo, J.W.; Mitani, S.; Sasaki, T.T.; Sukegawa, H.; Wen, Z.C.; Ohkubo, T.; Niizeki, T.; Inomata, K.; Hono, K. Large perpendicular magnetic anisotropy at Fe/MgO interface. Appl. Phys. Lett. 2013, 103, 192401. [CrossRef] Micromachines 2019, 10, 327 31 of 31

135. Lambert, C.H.; Rajanikanth, A.; Hauet, T.; Mangin, S.; Fullerton, E.E.; Andrieu, S. Quantifying perpendicular magnetic anisotropy at the Fe-MgO(001) interface. Appl. Phys. Lett. 2013, 102, 122410. [CrossRef] 136. Okabayashi, J.; Koo, J.W.; Sukegawa, H.; Mitani, S.; Takagi, Y.; Yokoyama, T. Perpendicular magnetic anisotropy at the interface between ultrathin Fe film and MgO studied by angular-dependent x-ray magnetic circular dichroism. Appl. Phys. Lett. 2014, 105, 122408. [CrossRef] 137. Bonell, F.; Lam, D.D.; Yoshida, S.; Takahashi, Y.T.; Shiota, Y.; Miwa, S.; Nakamura, T.; Suzuki, Y. Investigation of Au and Ag segregation on Fe(001) with soft X-ray absorption. Surf. Sci. 2013, 616, 125–130. [CrossRef] 138. Xiang, H.S.Q.; Al-Mahdawi, M.; Belmoubaric, M.; Kasai, S.; Sakuraba, Y.; Mitani, S.; Hono, K. Atomic layer number dependence of voltage-controlled magnetic anisotropy in Cr/Fe/MgAl2O4 heterostructure. In Proceedings of the Intermag2018, Singapore, FC-03, Marina Bay Sands Convention Centere, Singapore, 23–27 April 2018. 139. Pardede, I.; Kanagawa, T.; Ikhsan, N.; Murata, I.; Yoshikawa, D.; Obata, M.; Oda, T. A Comprehensive Study of Sign Change in Electric Field Control Perpendicular Magnetic Anisotropy Energy at Fe/MgO Interface: First Principles Calculation. IEEE Trans. Magn. 2019, 55, 1700104. [CrossRef] 140. Zhang, J.; Lukashev, P.V.; Jaswal, S.S.; Tsymbal, E.Y. Model of orbital populations for voltage-controlled magnetic anisotropy in transition -metal thin films. Phys. Rev. B 2017, 96, 014435. [CrossRef] 141. Nakamura, K.; Nomura, T.; Pradipto, A.M.; Nawa, K.; Akiyama, T.; Ito, T. Effect of heavy-metal insertions at Fe/MgO interfaces on electric-field-induced modification of magnetocrystalline anisotropy. J. Magn. Magn. Mater. 2017, 429, 214–220. [CrossRef] 142. Bonaedy, T.; Choi, J.W.; Jang, C.; Min, B.-C.; Chang, J. Enhancement of electric-field-induced change of magnetic anisotropy by interface engineering of MgO magnetic tunnel junctions. J. Phys. D Appl. Phys. 2015, 48, 225002. [CrossRef] 143. Nozaki, T.; Yamamoto, T.; Tamaru, S.; Kubota, H.; Fukushima, A.; Suzuki, A.Y.; Yuasa, S. Enhancement in the interfacial perpendicular magnetic anisotropy and the voltage-controlled magnetic anisotropy by heavy metal doping at the Fe/MgO interface. APL Mater. 2018, 6, 026101. [CrossRef] 144. Ikeura, T.; Nozaki, T.; Shiota, Y.; Yamamoto, T.; Imamura, H.; Kubota, H.; Fukushima, A.; Suzuki, Y.; Yuasa, S. Reduction in the write error rate of voltage-induced dynamic magnetization switching using the reverse bias method. Jpn. J. Appl. Phys. 2018, 57, 040311. [CrossRef] 145. Tamaru, S.; Yamamoto, T.; Nozaki, T.; Yuasa, S. Accurate calculation and shaping of the voltage pulse waveform applied to a voltage-controlled magnetic random access memory cell. Jpn. J. Appl. Phys. 2018, 57, 073002. [CrossRef] 146. Matsumoto, R.; Nozaki, T.; Yuasa, S.; Imamura, H. Voltage-Induced Precessional Switching at Zero-Bias Magnetic Field in a Conically Magnetized Free Layer. Phys. Rev. Appl. 2018, 9, 014026. [CrossRef] 147. Lee, J.-W.; Jeong, J.-R.; Shin, S.-C.; Kim, J.; Kim, S.-K. Spin-reorientation transitions in ultrathin Co films on Pt(111) and Pd(111) single-crystal substrates. Phys. Rev. B 2002, 66, 172409. [CrossRef] 148. Stamps, R.L.; Louail, L.; Hehn, M.; Gester, M.; Ounadjela, K. Anisotropies, cone states, and stripe domains in Co/Pt multilayers. J. Appl. Phys. 1997, 81, 4751–4753. [CrossRef] 149. Kisielewski, M.; Maziewski, A.; Tekielak, M.; Ferré, J.; Lemerle, S.; Mathet, V.; Chappert, C. Magnetic anisotropy and magnetization reversal processes in Pt/Co/Pt films. J. Magn. Magn. Mater. 2003, 260, 231–243. [CrossRef] 150. Shaw, J.M.; Nembach, H.T.; Weiler, M.; Silva, T.J.; Schoen, M.; Sun, J.Z.; Worledge, D.C. Perpendicular

Magnetic Anisotropy and Easy Cone State in Ta/Co60Fe20B20/MgO. IEEE Magn. Lett. 2015, 6, 3500404. [CrossRef] 151. Fu, Y.; Barsukov, I.; Li, J.; Gonçalves, A.M.; Kuo, C.C.; Farle, M.; Krivorotov, I.N. Temperature dependence of perpendicular magnetic anisotropy in CoFeB thin films. Appl. Phys. Lett. 2016, 108, 142403. [CrossRef] 152. Park, K.-W.; Park, J.-Y.; Baek, S.-H.C.; Kim, D.-H.; Seo, S.-M.; Chung, S.-W.; Park, B.-G. Electric field control of magnetic anisotropy in the easy cone state of Ta/Pt/CoFeB/MgO structures. Appl. Phys. Lett. 2016, 109, 012405. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).